CHAPTER 1
The purpose of this
chapter is to provide a basic introduction to modelling and
simulation. This chapter will serve to establish the basic definitions for terms that will be
used throughout the technical reference. In addition, emphasis will
be placed on the advantages of simulation.
When speaking about
modelling and simulation, the following terms are often used:
·
System
·
Experiment
·
Model
·
Simulation
System
A system is a set of
interdependent components that are united to perform a specified
function.
|
In a general sense,
the notion of a system may be defined as a collection of various
structural and non-structural elements which are interconnected and
organized to achieve some specified objective by the control and
distribution of material resources, energy and information. (Smith
et al., 1983)
One of the basic
aspects of a system is that it can be controlled and observed. Its
interactions with the environment fall into two categories:
1.
Variables generated by the environment that influence the behaviour
of the system (called inputs).
2.
Variables that are determined by the system that in turn influence
the behaviour of the environment (called outputs).
Accordingly, a
system is a potential source of data in that inputs can be defined
and observation of the behaviour of the system can be made.
Experiment
An experiment is
the process of extracting data from a system through manipulation
of the inputs.
Experimentation is
probably the single most important concept of a system; for it is
through experimentation that we develop a better understanding of
it. Experimentation implies that two basic properties of a system
are being used:
1.
Controllability, and
2.
Observability
To perform an
experiment implies the application of a set of external conditions
to the inputs of a system (i.e. the accessible inputs) and observe
the reaction of the system by recording the behaviour of the
outputs (i.e. the accessible outputs). This is where some of the
advantages of a system begin to appear. One of the major advantages
of experimenting with a "simulated" system as opposed to the
"actual" or "real" system, is that real systems are usually under
the influence of a large number of additional inaccessible inputs
(i.e., disturbances) and that a number of useful outputs may not be
available through measurement (i.e., they are internal states of
the system).
One of the major
motivations for simulation is that in the simulation world, all
inputs and outputs are accessible. This allows the execution of
simulations that lie outside the range of experiments that are
applicable to the real system.
Model
A model is an abstraction of a
system
|
One definition of a
model is: A model is an approximation of a system to which an
experiment can be applied to answer questions about the
system.
A model does
not imply a computer program. We should be clear to
distinguish between a model and a computer program. A model could
be a piece of hardware or simply an understanding of how a system
works. Models are often coded into computer programs.
Why is Modelling
Important?
Modelling means the
process of organizing knowledge about a given system. By performing
experiments, knowledge about a system is gathered. In the beginning
the knowledge is unstructured. By understanding the cause and
effect relationships and by placing observation in both a temporal
and spatial order, the knowledge gathered during the experiment is
organized. Thus, the system is better understood by the process of
modelling.
Simulation
Simulation is to a model what
experimentation is to a system
|
Again, many
different definitions exist for the term “simulation”. One of the
simplest definitions is: A simulation is an experiment performed
on a model Again, this does not imply that the simulation is
performed on a computer; however, the vast majority of engineering
simulations are performed using a computer program. A mathematical
simulation is a coded description of an experiment with a reference
point to the model to which this experiment is to be applied. The
goal is to be able to experiment with models as easily and
conveniently as with real systems. It is desired to be able to use
the simulation tools as easily as a control chart in the operation
of a facility.
While the scientist
is normally happy to observe and understand the world, that is,
creating a model of the world, the engineer (applied scientist)
wants to modify it to his/her advantage. While science is analysis,
the essence of engineering is control and design. Thus, simulation
can be used for analysis and for design.
Why is Simulation
Important?
Except by
experimenting with the real system, simulation is the only
technique available for the analysis of arbitrary system behaviour.
The typical scenario of scientific discovery is as follows:
1.
Perform an experiment on the real system and extract data to gather
knowledge (understanding of the cause and effect relationship of
the real world).
2.
Postulate a number of hypotheses related to the data.
3.
Simplify the problem to help make the analysis tractable.
4.
Perform a number of simulations with different experimental
parameters to verify that the simplifying assumptions are
justified.
5.
Analyze the system, verify the hypotheses and draw conclusions
6.
Simulations are performed to draw conclusions.
Simulation Tools
A wide variety of
simulation tools are available to help you in this task. Assuming
the reader is particularly interested in the dynamic modelling of
wastewater treatment, the tools appropriate for this task are
emphasized.
The process of
dynamic modelling of facilities involves the solution of thousands
of coupled nonlinear ordinary differential equations. The
formulation and solution of this type of problem is facilitated
through the use of Continuous Simulation Languages (CSL). CSLs date
back to the late 1960s when IBM introduced the language called CSMP
(Continuous System Modelling Program). Of the number of very
specialized simulation languages that are available, GPS-X uses
ACSL for conducting simulations.
Mathematical models
assist in developing a thorough understanding of the behaviour of a
system and in evaluating various system operating strategies. A
proposed system can be evaluated without building it. A costly or
unsafe system can be experimented with by using a model rather than
disturbing the real system.
One of the most
important ways to check the
operation of a wastewater treatment plant, the consistency of the
analytical procedures and the integrity of the mathematical model
is to perform mass balances around the system for the different
compounds. This task is not always simple as components transform
into other substances, bacterial cells grow, respire and decay.
With regard to the
organic substances, a commonly measurable parameter is the Chemical
Oxygen Demand (COD). We could measure organic carbon as Total Organic Carbon (TOC) in the
plant, but we would miss the fraction which was removed in the form
of CO2 gas after oxidation. It is difficult to
determine the oxygen
requirement based on TOC, as different substances require different
amounts of oxygen depending on their chemical composition. The
influent wastewater is truly a non-homogenous mixture in this respect.
We could measure
the 5-day Biochemical Oxygen Demand (BOD5) and
suspended solids as most plants in North America do. Suspended solids have the same
problem as TOC with respect to oxidation. BOD5
seems to give relevant information, but it is inappropriate for
continuous monitoring, and
the accuracy of the results is not comparable to other analytical
methods. BOD5 measures only the part of organics
which were used for respiration in the BOD test during 5 days, and
does not give information about the amount converted into bacterial
cells. Ultimate BOD (BODu) corrects this problem
but the analytical time and sometimes the accuracy is unacceptable.
The BOD test completely ignores a very important fraction of the
influent wastewater (inert particulates), which contributes in a
major way to excess sludge production.
COD overcomes the
above-mentioned problems. It can be automated and measures all
organic fractions of the wastewater. The sludge COD can also be
easily determined. COD measures all organics in oxygen equivalent; that is the electron donating capacity of the
organic matter. This way it provides a direct link between organic
load and aeration requirement. The yield constant is truly constant
only if expressed in COD units. Mass balance is easy to establish
with COD in a non-nitrifying plant: in steady-state, the influent
COD must equal the effluent COD plus the COD of the wasted sludge,
plus the oxygen consumed in the degradation of organic matter.
It is for this reason that the International Association on
Water Quality (IAWQ) committee selected and endorses the
use of COD as a measure of organic parameter in simulation of
activated sludge plants.
Data Requirements
For modelling
purposes, each unit process/operation is represented by a process
model (mathematical model) that reflects the dynamic behaviour of
that particular process. One of the main features of GPS-X is that
it is model-independent, meaning that GPS-X is not limited to a
specific process model. Accordingly, a variety of modelling
approaches (process
models) are available within GPS-X to handle a specific unit
operation or unit process. For example, the activated sludge
process can be modelled using any one of the following GPS-X
activated sludge process models:
·
IAWQ Task Group models of the activated sludge process (Henze et
al.,1987a; Henze et al., 1994; Henze et al.,
1998)
·
The general (bio-P) model (Dold, 1990, Barker and Dold, 1997)
·
Extended IAWQ (Mantis), described in Mantis Model
(MANTIS) section of Chapter 6)
·
Comprehensive plant-wide model developed by Hydromantis
(Mantis2/Mantis3)
Consequently, a
general calibration/verification approach to GPS-X must be broadly
defined. The calibration requirements of individual process models
are established based on the nature of each model (i.e., its
mechanistic basis). Alternatively, modellers may need to refer to
the original literature reference to assess the calibration
requirements of a particular model in more detail.
Each
calibration/verification study follows the same general principles.
Accordingly, the purpose of this section is to provide some
guidelines pertaining to the calibration of the models to full-scale wastewater
treatment plants. The most popular process models have been
selected for illustration purposes, including the IAWQ Task Group
Activated Sludge Model No. 1 (Henze et al., 1987a) and
layered settler model developed by Hydromantis (Takács et
al., 1991).
Overview of Data
Requirements
In general,
modelling of large-scale wastewater treatment plants requires that
an extensive number of plant and model parameters be assessed. Many
parameters can be measured directly, while others are based on
experimental data taken from the literature. Those parameters that
cannot be measured directly or estimated from the literature are
usually determined using nonlinear dynamic optimization techniques
based on actual plant records and/or experimental data collected at
the plant or in the lab. It is recognized that the reliability of
the calibrated model degrades with increasing numbers of
mathematically optimized parameters.
Data requirements
fall into one of the following categories:
1.
Physical plant data, including: Process flow sheet (flow
lines, channels, recycle lines, by-passes, etc.); Flow
pattern (plug flow, Continuously Stirred Tank Reactor (CSTR),
etc.); Sludge collection and withdrawal locations (location,
how? when? etc.); Dimensions of the various reactors
(length, width, depth).
2.
Operational plant data, including: Flow, Control variables
(independent variables), and Responsive variables (dependent
variables).
3.
Influent wastewater characteristics, including: Basic water
quality parameters, influent organic fractions, and influent
nitrogen fractions.
4.
Kinetic and stoichiometric model parameters for organic,
nitrogenous and phosphoric compounds and settling parameters
(primary and secondary).
5.
Some of these data and/or parameters vary in the course of a day
(i.e. subject to dry-weather diurnal variations or during a storm
even), while others remain relatively constant.
Elements of this
data group are generally easy to obtain from plant blueprints and
operation manuals. It should be remembered that the physical volume
of a reactor is only an approximation of the active or operational
volume of the unit. In a well-designed system the effect of
dead-space and hydraulic short-circuiting is normally minimal. In other
cases it may be necessary to determine the true hydraulic
characteristics of a particular unit process, as in the case of a
quasi-plug flow aeration tank. In this case, a dye-test is normally
required, as the number of CSTRs becomes a model parameter.
System Configuration
The General Purpose
Simulator can handle practically any flow scheme. Based on our
experience it is very important to identify as closely as possible
the hydraulic characteristics of a plant, including plant
by-passes, overflows, flow splits and combiners, proportional,
constant or SRT driven sludge wastage, etc. Parallel trains,
multiple units and plug flow systems are easily simulated, but
should be simplified where possible (unless the required supporting
data required for calibration is available).
Flow Control Variables
(Independent Variables)
This is an
important data group. For example, if the aeration capacity is not
known (or cannot be estimated from the aerator power or other
means), then the correct dissolved oxygen (DO) level can be set by
either changing the KLa or some stoichiometric or
kinetic parameters (yield coefficient, growth rate, etc.). This
makes the correct estimation of those parameters difficult.
Similarly, model parameters having a strong effect on the aeration
tank Mixed Liquor Suspended Solids (MLSS) are difficult to estimate
when the wastage rate is not known.
Activated Sludge Response
Variables
MLSS, Volatile
Suspended Solids (VSS), COD of the mixed liquor, DO, and Oxygen
Uptake Rate (OUR) are required to calibrate the activated sludge
portion of the model. Refer to above section (p. 22) for a
discussion on the importance of COD for this chapter (Why COD
is Important to Know). In general, the stoichiometry of the
mixed liquor (% VSS and COD/MLSS) is relatively constant over
time and can be assessed occasionally during the course of a
calibration/verification study, e.g., on a monthly or bi-weekly
basis. However, the other parameters are generally dynamic,
following the diurnal patterns of the plant.
It is important to be able to perform a solids
mass balance around the system. Accordingly, the sludge blanket
height (and preferably the solids concentration profile) and
underflow solids concentration are required to calibrate the
settler portion of the model.
Primary and Final Effluent
Response Variables
Water quality
constituents such as BOD5 (inhibited), Total
Suspended Solids (TSS), Total Kjeldahl Nitrogen (TKN), ammonia
(NH3) and nitrates (NO3) are
necessary for the calibration of the various unit processes. For
example, BOD (in lack of COD) is used to calibrate and verify the
carbonaceous component of the IAWQ activated sludge model, while
suspended solids measurements can be useful in identifying the
settling parameters of Hydromantis' layered settler model. The
nitrogenous compounds are needed to calibrate the
nitrification-denitrification component of the model.
Basic Parameters
Basic influent
wastewater characteristics such as BOD5,
BODu, COD, TSS, VSS, and TKN are important to
know in that they allow us to establish mass balances across the
system. The biochemical oxygen demand (BOD) provides only partial
information on the influent organic load (see Why COD is
Important to Know). COD measurements are not readily
available in some wastewater treatment plants. In this case, the
BOD5/BODu ratio can better
estimate the influent organic load. Suspended solids, influent VSS
and BOD together, can be used to determine the different influent
organic fractions, which are critical for the proper use of the
IAWQ activated sludge model, as discussed in GPS-X
Objects. Influent TKN is generally more useful than ammonia
concentration alone
Organic Compounds
The IAWQ activated
sludge model contains a large number of stoichiometric and kinetic
parameters, which describe the degradation of organic matter in the
activated sludge process (Henze et al., 1987a). Some of the
analytical tests are laborious and are not discussed here. Many of
the default model parameters
can be used with a high degree of confidence. Site-specific model
parameters include the maximum growth rate and the yield
coefficient of the heterotrophs. If the data described in the
previous sections are known (e.g., sludge wastage rate and
wastewater influent fractions), it is relatively easy to optimize
the maximum growth rate and yield coefficient of the heterotrophs
to match the measured MLSS, sludge production, and oxygen uptake
rate.
Nitrogenous Compounds
Based on our
experience, the most important parameter to calibrate in the IAWQ
model is the autotrophic
growth rate. It is possible to calibrate this parameter using field
ammonia and nitrate data, if:
1.
The plant is not overloaded, i.e. the plant is at least partially
nitrifying; or
2.
The plant is not seriously under loaded. In such a case, almost any
value of the growth rate constant (typically between 0.2-0.5
d-1) will provide complete nitrification.
The autotrophic
growth rate is easier to identify in a partially nitrifying plant.
Process start-up data (i.e., corresponding to a slowly developing
nitrifier population) can sometimes be used. Laboratory testing
(oxidation of an ammonia spike) is also a possibility.
Settling Characteristics
(Primary and Secondary)
The settling
velocity function in Hydromantis' layered settler model contains
five parameters, which have to be determined separately for the
primary and the secondary clarifiers. A preliminary version of the
model is described in detail elsewhere
(Takács et al., 1991). The model is based on the
use of a unified settling velocity equation described in the
chapter on sedimentation and flotation models. The parameters of
the settling velocity equation can be estimated from a combination
of experimental and numerical procedures.
A short summary of
the proposed experimental procedures is given below for each
parameter:
·
Minimum solids attainable – In general, this
parameter for final settlers is usually less than 10mg/L. For most
plants, xmin will be close to zero. A sludge sample is
allowed to settle for about two hours. The suspended solids
concentration of the supernatant is measured and equated to
xmin. Alternatively, xmin can be said to be equal to
the suspended solids concentration take from the final settler
under dry-weather flow conditions, when the hydraulic load to the
plant is minimal.
·
Maximum floc settling velocity parameter - Dilute the
activated sludge to 1 2 g/L, measure the settling
velocity of large individual floc particles in a batch test. In
general, no floc particle will settle faster than the settling
velocity of individual floc particles under quiescent
conditions.
·
Vesilind zone settling parameters – These two
parameters give the settling velocity of the sludge in the hindered
settling zone (exponential portion of the curve). They can be
determined through a series of column tests (Vesilind, 1968).
·
Flocculant settling parameter – If all the above
settling parameters are known, then this one is generally easy to
estimate by fitting the simulated effluent suspended solids
simulations to observed data.
Alternatively,
settling velocity model parameters can be estimated using a
time-series of influent and effluent (overflow and underflow)
suspended solids. The non-linear parameter optimization procedure
available in GPS-X can be used effectively in this particular
case.
A Typical Calibration
Event
In an ideal case
all the physical, operational and influent parameters are known for
the given wastewater treatment plant, while some of the most
important kinetic, stoichiometric and settling parameters are
experimentally determined. In such a case the modeller estimates
the missing parameters using defaults at the beginning, then
modifying those which need adjustment and observing the response of
several system output variables.
It is possible to
start with a steady-state
calibration, i.e., taking the dry weather days from a daily log of
the treatment plant and optimizing for the average of these values.
Averages, which contain high flow periods (typical monthly or
yearly averages), should not be used for steady-state
calibration.
Dynamic calibration
should follow with typical diurnal data or selected high
disturbance (storm flow) events. The larger the scale of the
disturbance between reasonable limits, the more sensitive the
calibration procedure will be. Hydraulic shocks are usually ideal
for settler calibrations, while diurnal data, process start-up, or
recovery is better for calibration of organic degradation and
nitrification.
One fully
documented event gives reasonable confidence for the given
conditions (flow, temperature, influent composition, etc.). If the
model is to be used under varying conditions, the above procedure
has to be repeated accordingly (i.e., winter, summer, dry weather,
wet weather, etc.). Verification means simulating a dynamic event
with a given calibrated set of parameters, without modifying those,
and finding reasonable accordance of simulated data with the
measurements.
A few or several
may be missing from the physical, operational and influent group.
This does not make calibration/verification impossible, as the
interdependencies in a treatment plant are complex and default
values are relatively well known. Reasonable estimation of
unknown parameters is sometimes possible. In the case of
underflows, the solids mass distribution between the aeration tank
and the settler depends largely on the recycle flow. Knowing
aeration tank MLSS and underflow concentration (maybe sludge
blanket height), the missing value can be recreated by numerically
fitting these variables as a function of the underflow. However,
with increasing number of such optimized parameters the confidence
in the predictions of the model erodes.
CHAPTER
2
An Object
refers to the icon, which appears on the GPS-X Drawing Board
when selected from the Process Table. An object is a
symbolic representation of a unit process without any process model
associated. There are several items associated with the object:
1.
The hydraulic configuration;
2.
Physical attributes;
3.
Operational attributes;
4.
Output Variables;
5.
Stream labels; and,
6.
Sources.
The hydraulic
configuration refers to the number of connection points on an
object and the operation of these connection points; that is,
whether the connection point is incoming or outgoing. The
parameters and stream labels are under the Parameters
sub-menu and Labels... sub-menu respectively. Additional
sub-menu items are found under the Parameters menu, which
are process model and library specific. The Output
Variablesmenu item is used to select model variables for
display on a graph, while the Sources menu item displays the
object number that acts as a source for some of the
Parameters.
Not all menus are
active. For example, until the model is chosen for an object, the
Parameters and Initialization menus (if they exist
for the object) are not active and appear greyed out. Once the
model has been specified, these menus will become active.
Common Properties
Before describing
each object that appears on the GPS-X Process Table, an
outline of the properties common to some objects is presented. The
objects in the Process Table can be described as either
having volume or not. There are some deviations from this general
rule, but they can be ignored for now. For example, the control
splitter object does not have any volume (called “zero volume”),
while the equalization basin does have volume. The zero volume
objects do not have any dilution or residence time while the
objects with volume do.
Objects with volume
have one or more influent connection points and one or more
effluent connection points. For example, the settler objects have
one influent connection point plus three effluent connection
points, while the PLUG FLOW-TANK(2)object has three influent
connection points and two effluent connection points. The effluent
connection point(s) will have an overflow connection plus one or
more pump connections. The overflow is located at the upper right
hand corner of the object (as oriented in the Process
Table), while any additional output connections are located on
the right hand side or bottom of the object. The additional output
connections (either called pump or underflow) simulate a constant
or variable flow pump so that a flow rate can be specified (up to
the maximum pump capacity).
The volume of fluid
in the tank depends on the net influent - effluent flow. If this is
a negative value, then the volume of the tank will decrease until
it reaches zero. At this point the effluent will equal the
influent, regardless of the pump flow set. If the net
influent-effluent flow is positive, the tank volume will increase
until the maximum (specified by the user). At this point, the tank
begins to overflow, so that the effluent flow (sum of the overflow
and pump flows) will be equal to the influent flow. The effluent
flow over and above the effluent pumped flow rates will leave
through the overflow connection point. If the net influent-effluent
flow is zero, then the volume will not change.
The mass balance
for variable volume tanks is shown in Equations 2.1-2.4:
Equation 2.1
Equation 2.2
Equation 2.3
Equation 2.4
where:
Qin =
influent flow rate (m3/d)
Qout = effluent
flow rate (m3/d)
Cin = influent
concentration (mg/L)
C =
effluent concentration (mg/L)
r
= rate of reaction (mg/L/d)
V =
liquid volume (m3)
t
= time (d)
Another feature of the objects with volume is their initial volume.
When a simulation begins, the user can specify what volume the tank
initially has through the use of a logical variable called
start with full tank, located under
Initial Conditions > Initial volume. If this
logical switch is true, then the tank will be full at the beginning
of the simulation.
If this logical variable is false, then the user can specify the
initial reactor volumeat the start of the simulation. As a
consequence of this, the user can specify the starting volume as
full by two ways:
1.
Either setting the logical variable as “true”; or
2.
By setting the variable as “false” and manually specifying the
starting volume as equal to the maximum tank volume.
From Equation
2.1, the concentration of a conservative material in
objects with volume relative to its influent concentration can be
calculated as shown in the following equations:
Equation 2.5
where:
X = conservative
component concentration (g/m3)
Q = flow rate
(m3/d)
t =
time (d)
Qin = influent
stream
Qout = effluent
stream
Xout = effluent
concentration (g/m3)
At steady-state,
the time derivatives are zero and Equation 2.5
becomes:
Equation 2.6
If the SRT is defined as:
Equation 2.7
where:
SRT = solids retention time (d)
V = volume
(m3)
Combining Equation 2.6 & Equation
2.7 gives:
Equation 2.8
The hydraulic residence time (HRT) is defined as:
Equation 2.9
Combining Equation 2.8 & Equation
2.9 gives:
Equation 2.10
This equation shows the ration of the concentration of the
conservative component in the object to its concentration in the
influent. At steady-state, it is directly proportional to the
SRT/HRT value.
CHAPTER 3
A library in GPS-X
is a collection of wastewater process models using a set of basic
wastewater components, or state variables. The term state
variable refers to the basic variables that are continuously
integrated over time. The composite variables are those
variables that are calculated from (or composed of) the state
variables. In discussing the state variables for the different
libraries listed below, volume is not explicitly explained as a
state variable as it is common to all libraries. The relationships
presented in this chapter between the state and composite variables
are used in every connection point of the plant layout.
NOTE:
In GPS-X, BOD refers to the carbonaceous BOD5
(CBOD) unless otherwise stated. This is to distinguish the
oxygen demand for organic carbon removal from the oxygen demand for
ammonia oxidation. The values of these two analyses for the same
sample can be considerably different.
Nine libraries are
available for GPS-X:
·
Comprehensive – Carbon,
Nitrogen, Phosphorus, pH (MANTIS2LIB)
·
Selenium and Sulfur (MANTIS2SLIB)
·
Carbon Footprint – Carbon,
Nitrogen, Phosphorus, pH (MANTIS3LIB)
·
Process Water Treatment Library
(PROCWATERLIB)
·
Petrochemical Wastewater
LIBRARY (MANTISIWLIB
·
Carbon – Nitrogen (CNLIB)
·
Carbon – Nitrogen – Industrial
Pollutant (CNIPLIB)
·
Carbon – Nitrogen – Phosphorus (CNPLIB)
·
Carbon – Nitrogen – Phosphorus
– Industrial Pollutant (CNPIPLIB)
NOTE:
MANTIS3LIB is available to those who have purchased the Carbon
Footprint Library add-on for their GPS-X license.
Fifty-two (52) state variables are available in the Comprehensive
Model (MANTIS2LIB) library. (Table 3‑1)
Table 3‑1 -
Comprehensie Model (MANTIS2LIB) Library State Variables
|
State Variables
|
GPS-X Cryptic Symbols
|
Units
|
1.
|
Dissolved
oxygen
|
so
|
gO2/m3
|
2.
|
Soluble inert
organic
|
si
|
gCOD/m3
|
3.
|
Colloidal organic
substrate
|
scol
|
gCOD/m3
|
4.
|
Fermentable
substrate
|
ss
|
gCOD/m3
|
5.
|
Acetate
|
sac
|
gCOD/m3
|
6.
|
Propionate
|
spro
|
gCOD/m3
|
7.
|
Methanol
|
smet
|
gCOD/m3
|
8.
|
Dissolved
hydrogen
|
sh2
|
gCOD/m3
|
9.
|
Dissolved
methane
|
sch4
|
gCOD/m3
|
10.
|
Dissolved inorganic
carbon
|
stic
|
gC/m3
|
11.
|
Soluble organic
nitrogen
|
snd
|
gN/m3
|
12.
|
Ammonia
nitrogen
|
snh
|
gN/m3
|
13.
|
Nitrite
nitrogen
|
snoi
|
gN/m3
|
14.
|
Nitrate
nitrogen
|
snoa
|
gN/m3
|
15.
|
Dissolved
nitrogen
|
sn2
|
gN/m3
|
16.
|
Ortho-phosphate
|
sp
|
gP/m3
|
17.
|
Dissolved
calcium
|
sca
|
gCa/m3
|
18.
|
Dissolved
magnesium
|
smg
|
gMg/m3
|
19.
|
Dissolved
potassium
|
spot
|
gK/m3
|
20.
|
Dissolved
cation
|
scat
|
eq/m3
|
21.
|
Dissolved
anion
|
sana
|
eq/m3
|
22.
|
Inert Particulate
|
xi
|
gCOD/m3
|
23.
|
Un-biodegradable cell decay material
|
xu
|
gO2/m3
|
24.
|
Slowly biodegradable organics
|
xs
|
gCOD/m3
|
25.
|
PHA accumulated in PAO
|
xbt
|
gCOD/m3
|
26.
|
Heterotrophic biomass
|
xbh
|
gCOD/m3
|
27.
|
Phosphate accumulating biomass
|
xbp
|
gCOD/m3
|
28.
|
Ammonia oxidizer
|
xbai
|
gCOD/m3
|
29.
|
Nitrite oxidizer
|
xbaa
|
gCOD/m3
|
30.
|
Anammox biomass
|
xbax
|
gCOD/m3
|
31.
|
Methylotrophic biomass
|
xmet
|
gCOD/m3
|
32.
|
Fermentative biomass
|
xbf
|
gC/m3
|
33.
|
Acetogen
|
xbpro
|
gN/m3
|
34.
|
Acetate methanogens
|
xbacm
|
gN/m3
|
35.
|
Hydrogen methanogens
|
xbh2m
|
gN/m3
|
36.
|
Nitrogen in slowly deg. organics
|
xns
|
gN/m3
|
37.
|
Phosphorous in slowly deg. organics
|
xps
|
gN/m3
|
38.
|
Poly-phosphate accumulated in PAO
|
xpp
|
gP/m3
|
39.
|
Particulate inert inorganic
|
xii
|
gCa/m3
|
40.
|
Aluminum hydroxide
|
xaloh
|
gMg/m3
|
41.
|
Aluminum phosphate
|
xalpo4
|
gK/m3
|
42.
|
Iron hydroxide
|
xfeoh
|
eq/m3
|
43.
|
Iron phosphate
|
xfepo4
|
eq/m3
|
44.
|
Calcium carbonate
|
xcaco3
|
gCOD/m3
|
45.
|
Calcium phosphate
|
xcapo4
|
gO2/m3
|
46.
|
Magnesium hydrogen phosphate
|
xmghpo4
|
gCOD/m3
|
47.
|
Magnesium carbonate
|
xmgco3
|
gCOD/m3
|
48.
|
Ammonium magnesium
phosphate(struvite)
|
xmgnh4po4
|
gCOD/m3
|
49.
|
Soluble component "a"
|
sza
|
gCOD/m3
|
50.
|
Soluble component "b"
|
szb
|
gCOD/m3
|
51.
|
Particulate component "a"
|
xza
|
gCOD/m3
|
52.
|
Particulate component "b"
|
xzb
|
gCOD/m3
|
Comprehensive Model
Library (MANTIS2SLIB)
Seventy-two (72)
state variables are available in the Sulphur and Selenium
(MANTIS2SLIB) library. (Table 3‑2)
Table 3‑2 -
Comprehensive Model (MANTIS2SLIB) Library State Variables
|
State Variables
|
GPS-X Cryptic Symbols
|
Units
|
1.
|
dissolved oxygen
|
so
|
gO2/m3
|
2.
|
dissolved hydrogen gas
|
sh2
|
gCOD/m3
|
3.
|
dissolved dinitrogen gas
|
sn2
|
gN/m3
|
4.
|
dissolved methane
|
sch4
|
gCOD/m3
|
5.
|
readily degradable soluble substrate
|
ss
|
gCOD/m3
|
6.
|
acetate
|
sac
|
gCOD/m3
|
7.
|
propionate
|
spro
|
gCOD/m3
|
8.
|
methanol
|
smet
|
gCOD/m3
|
9.
|
colloidal substrate
|
scol
|
gCOD/m3
|
10.
|
soluble inert material
|
si
|
gCOD/m3
|
11.
|
slowly biodegradable substrate 1
|
xs1
|
gCOD/m3
|
12.
|
slowly biodegradable substrate 2
|
xs2
|
gCOD/m3
|
13.
|
slowly biodegradable substrate 3
|
xs3
|
gCOD/m3
|
14.
|
poly-hydroxy alkanoates in PAO
|
xbt
|
gCOD/m3
|
15.
|
unbiodegradable cell products
|
xu
|
gCOD/m3
|
16.
|
particulate inert material
|
xi
|
gCOD/m3
|
17.
|
ammonia nitrogen
|
snh
|
gN/m3
|
18.
|
nitrite
|
snoi
|
gN/m3
|
19.
|
nitrate
|
snoa
|
gN/m3
|
20.
|
soluble organic nitrogen
|
snd
|
gN/m3
|
21.
|
nitrogen in slowly biodegradable
substrate
|
xns
|
gN/m3
|
22.
|
selenate selenium
|
sselna
|
gSe/m3
|
23.
|
selenite selenium
|
sselni
|
gSe/m3
|
24.
|
elemental selenium
|
xse0
|
gSe/m3
|
25.
|
soluble sulfide sulfur
|
stssul
|
gS/m3
|
26.
|
sulfate sulfur
|
sso4
|
gS/m3
|
27.
|
sulfite sulfur
|
sso3
|
gS/m3
|
28.
|
elemental sulfur
|
xsul0
|
gS/m3
|
29.
|
particulate heavy metal sulfide
|
xhmes
|
gS/m3
|
30.
|
ortho-phosphate
|
sp
|
gP/m3
|
31.
|
phosphorus in slowly biodegradable
substrate
|
xps
|
gP/m3
|
32.
|
poly-phosphate in PAO
|
xpp
|
gP/m3
|
33.
|
heterotrophic biomass
|
xbh
|
gCOD/m3
|
34.
|
ammonia oxidizer biomass
|
xbai
|
gCOD/m3
|
35.
|
nitrite oxidixer biomass
|
xbaa
|
gCOD/m3
|
36.
|
phosphate accumulating biomass
|
xbp
|
gCOD/m3
|
37.
|
selenium-reducing biomass
|
xbsel
|
gCOD/m3
|
38.
|
methylotrophic selenium-reducing
biomass
|
xbsemet
|
gCOD/m3
|
39.
|
propionate degrading sulfate-reducing
biomass
|
xbsr1
|
gCOD/m3
|
40.
|
hydrogen utilizing sulfate-reducing
biomass
|
xbsr2
|
gCOD/m3
|
41.
|
acetate utilizing sulfate-reducing
biomass
|
xbsr3
|
gCOD/m3
|
42.
|
methanol utilizing sulfate-reducing
biomass
|
xbsr4
|
gCOD/m3
|
43.
|
sulfur-oxiding biomass
|
xbsox
|
gCOD/m3
|
44.
|
fermenting biomass
|
xbf
|
gCOD/m3
|
45.
|
acetogenic biomass
|
xbpro
|
gCOD/m3
|
46.
|
acetoclastic methanogenic biomass
|
xbacm
|
gCOD/m3
|
47.
|
hydrogenotrophic methanogenic biomass
|
xbh2m
|
gCOD/m3
|
48.
|
methylotrophic biomass
|
xbmet
|
gCOD/m3
|
49.
|
methylotrophic methanogenic biomass
|
xbmem
|
gCOD/m3
|
50.
|
anammox biomass
|
xbax
|
gCOD/m3
|
51.
|
total soluble inorganic carbon
|
stic
|
gC/m3
|
52.
|
total soluble calcium
|
sca
|
gCa/m3
|
53.
|
soluble heavy metal
|
shme
|
gMe/m3
|
54.
|
total soluble magnesium
|
smg
|
gMg/m3
|
55.
|
total soluble potassium
|
spot
|
gK/m3
|
56.
|
other cation
|
scat
|
eq/m3
|
57.
|
other anion
|
sana
|
eq/m3
|
58.
|
inorganic inert particulate
|
xii
|
gSS/m3
|
59.
|
aluminum hydroxide
|
xaloh
|
gAl(OH)3/m3
|
60.
|
aluminum phosphate
|
xalpo4
|
gAlPO4/m3
|
61.
|
iron hydroxide
|
xfeoh
|
gFe(OH)3/m3
|
62.
|
iron sulfide
|
xfes
|
gFe(OH)3/m3
|
63.
|
iron phosphate
|
xfepo4
|
gFePO4/m3
|
64.
|
calcium carbonate
|
xcaco3
|
gCaCO3/m3
|
65.
|
calcium phosphate
|
xcapo4
|
gCaPO4/m3
|
66.
|
magnesium carbonate
|
xmgco3
|
gMgCO3/m3
|
67.
|
magnesium hydrogen phosphate
(newberyite)
|
xmghpo4
|
gMgHPO4.3H2O/m3
|
68.
|
magnesium ammonium phosphate
(struvite)
|
xmgnh4po4
|
gMgNH4PO4.6H2O/m3
|
69.
|
soluble component a
|
sza
|
notset
|
70.
|
soluble component b
|
szb
|
notset
|
71.
|
particulate component a
|
xza
|
notset
|
72.
|
particulate component b
|
xzb
|
notset
|
Carbon Footprint – Carbon,
Nitrogen, Phosphorus, pH (MANTIS3LIB)
Fifty-six (56) state variables are available in the Carbon
Footprint (MANTIS3LIB) Library. (Table
3‑3).
Table 3‑3 - Carbon
Footprint (MANTIS3LIB) Library State Variables
|
State Variables
|
GPS-X Cryptic Symbols
|
Units
|
1.
|
Dissolved
oxygen
|
so
|
gO2/m3
|
2.
|
Soluble inert
organic
|
si
|
gCOD/m3
|
3.
|
Colloidal organic
substrate
|
scol
|
gCOD/m3
|
4.
|
Fermentable
substrate
|
ss
|
gCOD/m3
|
5.
|
Acetate
|
sac
|
gCOD/m3
|
6.
|
Propionate
|
spro
|
gCOD/m3
|
7.
|
Methanol
|
smet
|
gCOD/m3
|
8.
|
Dissolved
hydrogen
|
sh2
|
gCOD/m3
|
9.
|
Dissolved
methane
|
sch4
|
gCOD/m3
|
10.
|
Dissolved inorganic
carbon
|
stic
|
gC/m3
|
11.
|
Soluble organic
nitrogen
|
snd
|
gN/m3
|
12.
|
Ammonia
nitrogen
|
snh
|
gN/m3
|
13.
|
Nitrite
nitrogen
|
snoi
|
gN/m3
|
14.
|
Nitrate
nitrogen
|
snoa
|
gN/m3
|
15.
|
Dissolved
nitrogen
|
sn2
|
gN/m3
|
16.
|
Nitric
oxide-Nitrogen
|
snrio
|
gN/m3
|
17.
|
Nitrous Oxide
|
snroo
|
gN/m3
|
18.
|
Hydroxylamine
|
snh2oh
|
gN/m3
|
19.
|
Nitrosyl
radical
|
snoh
|
gN/m3
|
20.
|
Ortho-phosphate
|
sp
|
gP/m3
|
21.
|
Dissolved
calcium
|
sca
|
gCa/m3
|
22.
|
Dissolved
magnesium
|
smg
|
gMg/m3
|
23.
|
Dissolved
potassium
|
spot
|
gK/m3
|
24.
|
Dissolved
cation
|
scat
|
eq/m3
|
25.
|
Dissolved
anion
|
sana
|
eq/m3
|
26.
|
Inert Particulate
|
xi
|
gCOD/m3
|
27.
|
Un-biodegradable cell decay material
|
xu
|
gCOD/m3
|
28.
|
Slowly biodegradable organics
|
xs
|
gCOD/m3
|
29.
|
PHA accumulated in PAO
|
xbt
|
gCOD/m3
|
30.
|
Heterotrophic biomass
|
xbh
|
gCOD/m3
|
31.
|
Phosphate accumulating biomass
|
xbp
|
gCOD/m3
|
32.
|
Ammonia oxidizer
|
xbai
|
gCOD/m3
|
33.
|
Nitrite oxidizer
|
xbaa
|
gCOD/m3
|
34.
|
Anammox biomass
|
xbax
|
g COD/m3
|
35.
|
Methylotrophic biomass
|
xmet
|
g COD/m3
|
36.
|
Fermentative biomass
|
xbf
|
g COD/m3
|
37.
|
Acetogen
|
xbpro
|
gCOD/m3
|
38.
|
Acetate methanogens
|
xbacm
|
gCOD/m3
|
39.
|
Hydrogen methanogens
|
xbh2m
|
gCOD/m3
|
40.
|
Nitrogen in slowly deg. organics
|
xns
|
gN/m3
|
41.
|
Phosphorous in slowly deg. organics
|
xps
|
gP/m3
|
42.
|
Poly-phosphate accumulated in PAO
|
xpp
|
gP/m3
|
43.
|
Particulate inert inorganic
|
xii
|
g/m3
|
44.
|
Aluminum hydroxide
|
xaloh
|
gAl(OH)3/m3
|
45.
|
Aluminum phosphate
|
xalpo4
|
gAlPO4/m3
|
46.
|
Iron hydroxide
|
xfeoh
|
gFe(OH)3/m3
|
47.
|
Iron phosphate
|
xfepo4
|
gFePO4/m3
|
48.
|
Calcium carbonate
|
xcaco3
|
gCaCO3/m3
|
49.
|
Calcium phosphate
|
xcapo4
|
gCa3(PO4)2/m3
|
50.
|
Magnesium hydrogen phosphate
|
xmghpo4
|
gMgHPO4.3H2O/m3
|
51.
|
Magnesium carbonate
|
xmgco3
|
gMgCO3/m3
|
52.
|
Ammonium magnesium
phosphate(struvite)
|
xmgnh4po4
|
gMgNH4PO4.6H2O
/m3
|
53.
|
Soluble component "a"
|
sza
|
notset
|
54.
|
Soluble component "b"
|
szb
|
notset
|
55.
|
Particulate component "a"
|
xza
|
notset
|
56.
|
Particulate component "b"
|
xzb
|
notset
|
Carbon – Nitrogen Library
(CNLIB)
Sixteen state
variables are available in the Carbon – Nitrogen library
(Table 3‑4)
Table 3‑4 –
Carbon – Nitrogen Library (CNLIB) State Variables
|
State Variables
|
GPS-X Cryptic Symbols
|
Units
|
1.
|
Soluble inert organics
|
si
|
gCOD/m3
|
2.
|
Readily biodegradable (soluble)
substrate
|
ss
|
gCOD/m3
|
3.
|
Particulate inert
organics
|
xi
|
gCOD/m3
|
4.
|
Slowly biodegr. (stored,
particulate) substrate
|
xs
|
gCOD/m3
|
5.
|
Active heterotrophic
biomass
|
xbh
|
gCOD/m3
|
6.
|
Active autotrophic
biomass
|
xba
|
gCOD/m3
|
7.
|
Unbiodegradable particulates from
cell decay
|
xu
|
gCOD/m3
|
8.
|
Cell internal storage
product
|
xsto
|
gCOD/m3
|
9.
|
Dissolved oxygen
|
so
|
gN/m3
|
10.
|
Nitrate and nitrite N
|
sno
|
gN/m3
|
11.
|
Free and ionized ammonia
|
snh
|
gN/m3
|
12.
|
Soluble biodegradable organic
nitrogen (in ss)
|
snd
|
gN/m3
|
13.
|
Particulate biodegr. organic
nitrogen (in xs)
|
xnd
|
gN/m3
|
14.
|
Dinitrogen
|
snn
|
gN/m3
|
15.
|
Alkalinity
|
salk
|
mole/m3
|
16.
|
Inert inorganic suspended
solids
|
xii
|
g/m3
|
Industrial Pollutant Library (CNIPLIB)
Forty-six (46)
state variables are available in the Industrial Pollutant library.
Sixteen (16) are pre-defined and thirty (30) are user-definable (15
soluble, 15 particulate). They are listed in Table
3‑5.
Table 3‑5 –
Industrial Pollutant (CNIPLIB) Library State Variables
|
State Variables
|
GPS-X Cryptic Symbols
|
Units
|
1.
|
Soluble inert organics
|
si
|
gCOD/m3
|
2.
|
Readily biodegradable (soluble)
substrate
|
ss
|
gCOD/m3
|
3.
|
Particulate inert organics
|
xi
|
gCOD/m3
|
4.
|
Slowly biodegr. (stored, particulate)
substrate
|
xs
|
gCOD/m3
|
5.
|
Active heterotrophic biomass
|
xbh
|
gCOD/m3
|
6.
|
Active autotrophic biomass
|
xba
|
gCOD/m3
|
7.
|
Unbiodegradable particulates from cell
decay
|
xu
|
gCOD/m3
|
8.
|
Cell internal storage product
|
xsto
|
gCOD/m3
|
9.
|
Dissolved oxygen
|
so
|
gO2/m3
|
10.
|
Nitrate and nitrite N
|
sno
|
gN/m3
|
11.
|
Free and ionized ammonia
|
snh
|
gN/m3
|
12.
|
Soluble biodegradable organic nitrogen (in
ss)
|
snd
|
gN/m3
|
13.
|
Particulate biodegradable organic nitrogen (in
xs)
|
xnd
|
gN/m3
|
14.
|
Dinitrogen
|
snn
|
gN/m3
|
15.
|
Alkalinity
|
salk
|
mole/m3
|
16.
|
Inert inorganic suspended solids
|
xii
|
g/m3
|
17.
|
Soluble component "a"
|
sza
|
notset
|
18.
|
Soluble component "b"
|
szb
|
notset
|
19.
|
Soluble component "c"
|
szc
|
notset
|
20.
|
Soluble component "d"
|
szd
|
notset
|
21.
|
Soluble component "e"
|
sze
|
notset
|
22.
|
Soluble component "f"
|
szf
|
notset
|
23.
|
Soluble component "g"
|
szg
|
notset
|
24.
|
Soluble component "h"
|
szh
|
notset
|
25.
|
Soluble component "i"
|
szi
|
notset
|
26.
|
Soluble component "j"
|
szj
|
notset
|
27.
|
Soluble component "k"
|
szk
|
notset
|
28.
|
Soluble component "l"
|
szl
|
notset
|
29.
|
Soluble component "m"
|
szm
|
notset
|
30.
|
Soluble component "n"
|
szn
|
notset
|
31.
|
Soluble component "o"
|
szo
|
notset
|
32.
|
Particulate component "a"
|
xza
|
notset
|
33.
|
Particulate component "b"
|
xzb
|
notset
|
34.
|
Particulate component "c"
|
xzc
|
notset
|
35.
|
Particulate component "d"
|
xzd
|
notset
|
36.
|
Particulate component "e"
|
xze
|
notset
|
37.
|
Particulate component "f"
|
xzf
|
notset
|
38.
|
Particulate component "g"
|
xzg
|
notset
|
39.
|
Particulate component "h"
|
xzh
|
notset
|
40.
|
Particulate component "i"
|
xzi
|
notset
|
41.
|
Particulate component "j"
|
xzj
|
notset
|
42.
|
Particulate component "k"
|
xzk
|
notset
|
43.
|
Particulate component "l"
|
xzl
|
notset
|
44.
|
Particulate component "m"
|
xzm
|
notset
|
45.
|
Particulate component "n"
|
xzn
|
notset
|
46.
|
Particulate component "o"
|
xzo
|
notset
|
Carbon – Nitrogen –
Phosphorus Library (CNPLIB)
Twenty-seven (27)
state variables are available in the Carbon – Nitrogen – Phosphorus
library (Table 3‑6)
Table 3‑6 –
Carbon – Nitrogen – Phosphorus (CNPLIB) Library State Variables
|
State Variables
|
GPS-X Cryptic Symbols
|
Units
|
1.
|
Soluble inert organics
|
si
|
gCOD/m3
|
2.
|
Readily biodegradable (soluble)
substrate
|
ss
|
gCOD/m3
|
3.
|
Particulate inert
organics
|
xi
|
gCOD/m3
|
4.
|
Slowly biodegr. (stored,
particulate) substrate
|
xs
|
gCOD/m3
|
5.
|
Active heterotrophic
biomass
|
xbh
|
gCOD/m3
|
6.
|
Active autotrophic
biomass
|
xba
|
gCOD/m3
|
7.
|
Unbiodegradable particulates from
cell decay
|
xu
|
gCOD/m3
|
8.
|
Dissolved oxygen
|
so
|
gO2/m3
|
9.
|
Nitrate and nitrite N
|
sno
|
gN/m3
|
10.
|
Free and ionized ammonia
|
snh
|
gN/m3
|
11.
|
Soluble biodegradable organic
nitrogen (in ss)
|
snd
|
gN/m3
|
12.
|
Particulate biodegradable organic
nitrogen (in xs)
|
xnd
|
gN/m3
|
13.
|
Polyphosphate accumulating
biomass
|
xbp
|
gCOD/m3
|
14.
|
Poly-hydroxy-alkanoates
(PHA)
|
xbt
|
gCOD/m3
|
15.
|
Stored polyphosphate
|
xpp
|
gP/m3
|
16.
|
Volatile fatty acids
|
slf
|
gCOD/m3
|
17.
|
Soluble phosphorus
|
sp
|
gP/m3
|
18.
|
Alkalinity
|
salk
|
mole/m3
|
19.
|
Dinitrogen
|
snn
|
gN/m3
|
20.
|
Soluble unbiodegradable organic
nitrogen (in si)
|
sni
|
gN/m3
|
21.
|
Fermentable readily biodegradable
substrate
|
sf
|
gCOD/m3
|
22.
|
Stored glycogen
|
xgly
|
gCOD/m3
|
23.
|
Stored polyphosphate
(releasable)
|
xppr
|
gP/m3
|
24.
|
Metal-hydroxides
|
xmeoh
|
g/m3
|
25.
|
Metal-phosphate
|
xmep
|
g/m3
|
26.
|
Cell internal storage
product
|
xsto
|
gCOD/m3
|
27.
|
Inert inorganic suspended
solids
|
xii
|
g/m3
|
CNP Industrial Pollutant
Library (CNPIPLIB)
Fifty-seven (57)
state variables are available in the CNP Industrial Pollutant
Library (CNPIPLIB). These include the 27 state variables from
CNPLIB, as well as 30 user‑definable variables (15 soluble and 15
particulate) as shown in Table 3‑7
Table 3‑7 –
CNP Industrial Pollutant (CNPIPLIB) Library State Variables
|
State Variables
|
GPS-X Cryptic Symbols
|
Units
|
1.
|
Soluble inert organics
|
si
|
gCOD/m3
|
2.
|
Readily biodegradable (soluble)
substrate
|
ss
|
gCOD/m3
|
3.
|
Particulate inert organics
|
xi
|
gCOD/m3
|
4.
|
Slowly biodegr. (stored, particulate)
substrate
|
xs
|
gCOD/m3
|
5.
|
Active heterotrophic biomass
|
xbh
|
gCOD/m3
|
6.
|
Active autotrophic biomass
|
xba
|
gCOD/m3
|
7.
|
Unbiodegradable particulates from cell
decay
|
xu
|
gCOD/m3
|
8.
|
Dissolved oxygen
|
so
|
gO2/m3
|
9.
|
Nitrate and nitrite N
|
sno
|
gN/m3
|
10.
|
Free and ionized ammonia
|
snh
|
gN/m3
|
11.
|
Soluble biodegradable organic nitrogen (in
ss)
|
snd
|
gN/m3
|
12.
|
Particulate biodegradable organic nitrogen (in
xs)
|
xnd
|
gN/m3
|
13.
|
Polyphosphate accumulating biomass
|
xbp
|
gCOD/m3
|
14.
|
Poly-hydroxy-alkanoates (PHA)
|
xbt
|
gCOD/m3
|
15.
|
Stored polyphosphate
|
xpp
|
gP/m3
|
16.
|
Volatile fatty acids
|
slf
|
gCOD/m3
|
17.
|
Soluble phosphorus
|
sp
|
gP/m3
|
18.
|
Alkalinity
|
salk
|
mole/m3
|
19.
|
Dinitrogen
|
snn
|
gN/m3
|
20.
|
Soluble unbiodegradable organic nitrogen (in
si)
|
sni
|
gN/m3
|
21.
|
Fermentable readily biodegradable
substrate
|
sf
|
gCOD/m3
|
22.
|
Stored glycogen
|
xgly
|
gCOD/m3
|
23.
|
Stored polyphosphate (releasable)
|
xppr
|
gP/m3
|
24.
|
Metal-hydroxides
|
xmeoh
|
g/m3
|
25.
|
Metal-phosphate
|
xmep
|
g/m3
|
26.
|
Cell internal storage product
|
xsto
|
gCOD/m3
|
27.
|
Inert inorganic suspended solids
|
xii
|
g/m3
|
28.
|
Soluble component "a"
|
sza
|
notset
|
29.
|
Soluble component "b"
|
szb
|
notset
|
30.
|
Soluble component "c"
|
szc
|
notset
|
31.
|
Soluble component "d"
|
szd
|
notset
|
32.
|
Soluble component "e"
|
sze
|
notset
|
33.
|
Soluble component "f"
|
szf
|
notset
|
34.
|
Soluble component "g"
|
szg
|
notset
|
35.
|
Soluble component "h"
|
szh
|
notset
|
36.
|
Soluble component "i"
|
szi
|
notset
|
37.
|
Soluble component "j"
|
szj
|
notset
|
38.
|
Soluble component "k"
|
szk
|
notset
|
39.
|
Soluble component "l"
|
szl
|
notset
|
40.
|
Soluble component "m"
|
szm
|
notset
|
41.
|
Soluble component "n"
|
szn
|
notset
|
42.
|
Soluble component "o"
|
szo
|
notset
|
43.
|
Particulate component "a"
|
xza
|
notset
|
44.
|
Particulate component "b"
|
xzb
|
notset
|
45.
|
Particulate component "c"
|
xzc
|
notset
|
46.
|
Particulate component "d"
|
xzd
|
notset
|
47.
|
Particulate component "e"
|
xze
|
notset
|
48.
|
Particulate component "f"
|
xzf
|
notset
|
49.
|
Particulate component "g"
|
xzg
|
notset
|
50.
|
Particulate component "h"
|
xzh
|
notset
|
51.
|
Particulate component "i"
|
xzi
|
notset
|
52.
|
Particulate component "j"
|
xzj
|
notset
|
53.
|
Particulate component "k"
|
xzk
|
notset
|
54.
|
Particulate component "l"
|
xzl
|
notset
|
55.
|
Particulate component "m"
|
xzm
|
notset
|
56.
|
Particulate component "n"
|
xzn
|
notset
|
57.
|
Particulate component "o"
|
xzo
|
notset
|
CHAPTER
4
In GPS-X, a group
of state variables (such as oxygen, heterotrophic biomass, nitrate,
ammonia, soluble substrate, particulate substrate, etc.) are
calculated for each connection point in the plant layout. These
state variables are the fundamental components that are acted upon
by the processes in the models in each library.
These particular
state variable components are not always easily measurable or
interpretable in practical applications. Therefore, a series of
composite variables are calculated from the state variables. The
composite variables combine the state variables into forms that are
typically measured, such as total suspended solids (TSS), BOD, COD
and Total Kjeldahl Nitrogen (TKN).
The way that the
composite variables are calculated from state variables changes
from library to library and to a great degree from model to
model.
Composite variables are
calculated from state variables with the use of
stoichiometric constants. These constants describe
the relationships between various states and composites, and depend
on the type of composite calculations used.
Nomenclature
In this chapter,
diagrams and tables are used to depict the relationships between
the state and composite variables
included in a library.
Box-and-arrow
Diagrams
The nomenclature
used in the box-and-arrows diagrams is explained in Figure
4‑1.
Figure 4‑1 – Diagram Nomenclature
The variables in
the boxes and above the connection lines are known (either
previously calculated or user input). The variables in
BOLD CAPITALS represent the composite variables which are to
be calculated. The connection line shows the direction of
calculation and always begins from a known or boxed variable.
Multiple lines converging to one unknown variable imply a summation
operator. In the example shown above, the variable Y1 is
calculated by multiplying the variable x1 by the
stoichiometry parameter k and summing it with variable
x2. If no stoichiometry parameter appears above the
connection line, it implies a default value of 1. When a broken
line circle is drawn on the lines, it indicates that the
stoichiometry parameters for these lines are model dependent.
In certain situations two or more calculated composite variables
are used to calculate an additional composite variable. For
example, Y3 is calculated by adding the calculated composite
variables of Y1 and Y2.
In addition to
diagrams which explain the general way composite variables are
calculated, composite variable tables are used to explain
model-specific calculations. The nomenclature used in the composite
variables tables is explained in Table 4‑1:
Table 4‑1 – Example Composite Variable
Calculations
|
SCOMP
|
XCOMP
|
TCOMP
|
sa
|
1
|
|
1
|
sb
|
ksb
|
|
ksb
|
xa
|
|
1
|
1
|
xb
|
|
kxb
|
kxb
|
The composite
variables being calculated are shown across the top of each column.
The state variables used in the calculations are shown down the
left side of the table. To calculate the composite variable, each
state variable is multiplied by the coefficient in the table for
that particular composite variable, and then summed down the
column.
For the example
data given in Table 4‑1, the calculations for
SCOMP, XCOMP, and TCOMP
are:
SCOMP = 1*sa + ksb*sb + 0*xa + 0*xb = sa +
ksb*sb
XCOMP = 0*sa + 0*sb + 1*xa + kxb*xb = xa +
kxb*xb
TCOMP = 1*sa + ksb*sb + 1*xa + kxb*xb = sa + ksb*sb +
xa + kxb*xb
Figure
4‑2 shows the relationship between the CNLIB state
variables and the TSS, BOD, and COD composite state variables.
Table 4‑2 illustrates the same composite variable
calculations in the tabular format.
Figure 4‑2 – Carbon – Nitrogen Library (CNLIB):
BOD, COD, and Suspended Solids Composite Variables and their
Relationship to the State Variables
Table 4‑2 –
CNLIB BOD, COD, and TSS Composite Variables (All Models)
|
SBODU
|
XBODU
|
BODU
|
SBOD
|
XBOD
|
BOD
|
SCOD
|
XCOD
|
COD
|
VSS
|
XISS
|
X
|
si
|
|
|
|
|
|
|
1
|
|
1
|
|
|
|
ss
|
1
|
|
1
|
fbod
|
|
fbod
|
1
|
|
1
|
|
|
|
xi
|
|
|
|
|
|
|
|
1
|
1
|
icv-1
|
|
icv-1
|
xs
|
|
1
|
1
|
|
fbod
|
fbod
|
|
1
|
1
|
icv-1
|
|
icv-1
|
xbh
|
|
1
|
1
|
|
fbod
|
fbod
|
|
1
|
1
|
icv-1
|
|
icv-1
|
xba
|
|
1
|
1
|
|
fbod
|
fbod
|
|
1
|
1
|
icv-1
|
|
icv-1
|
xu
|
|
|
|
|
|
|
|
1
|
1
|
icv-1
|
|
icv-1
|
xsto
|
|
1
|
1
|
|
fbod
|
fbod
|
|
1
|
1
|
icv-1
|
|
icv-1
|
xii
|
|
|
|
|
|
|
|
|
|
|
1
|
1
|
The COD composite variables are a sum of
state variables (where units are gCOD/m3). Soluble COD
(SCOD) is the sum of the soluble inert
organics (si) and readily biodegradable substrate
(ss), while particulate COD (XCOD) is the sum of the slowly
biodegradable substrate (xs), active heterotrophic biomass
(xbh), active autotrophic biomass (xba), cell
internal storage product (xsto), un-biodegradable
particulates from cell decay (xu), and particulate inert
organics (xi). The total COD (COD) is the sum
of the soluble and particulate COD.
The suspended solids composite variable (X) is
calculated from the particulate COD (XCOD) by
dividing it by the XCOD: VSS ratio (icv), which changes the units of the
XCOD to mgVSS/L, resulting in the volatile suspended
solids (VSS) composite variable. To calculate the
suspended solids composite variable (X), VSS
is added to particulate inert inorganic material
(XII). By default in the CN library, particulate
inert suspended solids (XISS) is equal to
xii.
The biochemical oxygen demand (BOD) composite
variables are calculated from the state variables. First, the
biodegradable state variables (the state variables that exert BOD,
which are in units of gO/m3) are summed to provide both
a particulate and a soluble ultimate BOD (XBODU,
SBODU). The sum of these components is the total
ultimate BOD measurement (BODU). To determine
BOD5, the calculated BODU is
multiplied by a stoichiometric fraction, fbod, which is the ratio
of BOD5:BODU.
In terms of variables containing nitrogen, the
calculation of the composite variables involves adding up various
state variables and multiplying other state variables by fractions
as appropriate (Figure
4‑3). The soluble total
Kjeldahl nitrogen (STKN) is the sum of the free and
ionized ammonia (snh), soluble biodegradable organic
nitrogen (snd), and (in asm3 only) the
nitrogen components of soluble substrate (ss) and soluble
inerts (si). The particulate total Kjeldahl nitrogen
(XTKN) is the particulate biodegradable organic
nitrogen (xnd), plus the nitrogen component of biomass
(xbh and xba), unbiodegradable cell products
(xu), particulate substrate (xs) and particulate
inerts (xi). Total Kjeldahl nitrogen (TKN) is
the sum of soluble (STKN) and particulate TKN
(XTKN). The total nitrogen (TN) is the
sum of the TKN and the nitrate nitrogen
(sno).
Figure 4‑3 – Carbon – Nitrogen Library:
Composite Variables and their Relationships to the State
Variables
The nitrogen
fractions of xbh, xba, xu, xi, ss and si vary from
model to model. The nitrogen composite variable relationships for
the mantis, asm1 and asm3 models are shown in
Table 4‑3, Table 4‑4, and Table
4‑5
Table 4‑3 –
CNLIB Nitrogen Composite Variables – MANTIS Model
|
STKN
|
XTKN
|
TKN
|
TN
|
sno
|
|
|
|
1
|
snh
|
1
|
|
1
|
1
|
snd
|
1
|
|
1
|
1
|
xnd
|
|
1
|
1
|
1
|
xbh
|
|
ibhn
|
ibhn
|
ibhn
|
xba
|
|
ibhn
|
ibhn
|
ibhn
|
xu
|
|
iuhn
|
iuhn
|
iuhn
|
xi
|
|
iuhn
|
iuhn
|
iuhn
|
Table 4‑4 –
CNLIB Nitrogen Composite Variables – ASM1 Model
|
STKN
|
XTKN
|
TKN
|
TN
|
sno
|
|
|
|
1
|
snh
|
1
|
|
1
|
1
|
snd
|
1
|
|
1
|
1
|
xnd
|
|
1
|
1
|
1
|
xbh
|
|
ixbn
|
ixbn
|
ixbn
|
xba
|
|
ixbn
|
ixbn
|
ixbn
|
xu
|
|
ixun
|
ixun
|
ixun
|
xi
|
|
ixun
|
ixun
|
ixun
|
Table 4‑5 - CNLIB
Nitrogen Composite Variables - ASM3 Model
|
STKN
|
XTKN
|
TKN
|
TN
|
sno
|
|
|
|
1
|
snh
|
1
|
|
1
|
1
|
si
|
insi
|
|
insi
|
insi
|
ss
|
inss
|
|
inss
|
inss
|
xbh
|
|
inbm
|
inbm
|
inbm
|
xba
|
|
inbm
|
inbm
|
inbm
|
xs
|
|
inxs
|
inxs
|
inxs
|
xi
|
|
inxi
|
inxi
|
inxi
|
snd
|
1
|
|
1
|
1
|
xnd
|
|
1
|
1
|
1
|
The relationship
between the state and composite variables are the same as those for
CNLIB (Figure 4‑3 and Table 4‑5).
Table 4‑3 through Table 4‑5 illustrates
the composite variable calculations for the mantis1,
asm1, and asm3 models in CNIPLIB.
Figure
4‑4 shows the relationships between CNPLIB state variables
and the TSS, BOD and COD composite variables. Table
4‑6 illustrates the same composite variable calculations in
tabular format.
As all the state
variables are in units of gCOD/m3, the COD composite
variables are simply a sum of the appropriate state variables. In
general, the soluble COD (SCOD) is a sum of the
soluble inert organics (si), volatile fatty acids
(slf), fermentable readily biodegradable substrate
(sf), and readily biodegradable substrate (ss). The
particulate COD (XCOD) is a sum of the slowly
biodegradable substrate (xs), active heterotrophic biomass
(xbh), active autotrophic biomass (xba),
polyphosphate accumulating biomass (xbp), stored glycogen
(xgly), poly-hydroxy-alkanoates (xbt),
unbiodegradable particulates from cell decay (xu), cell
internal storage product (xsto) and particulate inert
organics (xi). The total COD (COD) is the sum
of the soluble and particulate COD.
Figure 4‑4 – Carbon – Nitrogen – Phosphorus
Library: BOD, COD, and Suspended Solids Composite Variables and
their Relationship to the State Variables
The suspended
solids composite variable is calculated from the particulate COD
(XCOD) by dividing it by the XCOD:VSS ratio
(icv). This changes the units of the XCOD to
gVSS/m3, resulting in the composite variable for
volatile suspended solids (VSS). To calculate the
suspended solids composite variable (X), VSS is added
to the concentration of inert inorganic particulates
(XII), metal hydroxides (xmeoh) and metal
phosphates (xmep), and stored polyphosphate (multiplied by 3 to
change from gP/m3 to g/m3).
The biochemical
oxygen demand (BOD) composite variables are calculated from the
state variables. The biodegradable state variables (the state
variables that exert BOD, which are in units of gCOD/m3)
are summed to provide a particulate and a soluble ultimate BOD
(XBODU, SBODU). The sum of these
components is the total ultimate BOD measurement
(BODU). To determine BOD5, a
stoichiometric fraction, fbod, which is the ratio of
BOD5:BOD20, is multiplied by
the calculated BODU.
Table 4‑6 -
CNPLIB BOD, COD, and TSS Composite Variables
|
SBODU
|
XBODU
|
BODU
|
SBOD
|
XBOD
|
BOD
|
SCOD
|
XCOD
|
COD
|
VSS
|
XISS
|
X
|
ss
|
1
|
|
1
|
fbod
|
|
fbod
|
1
|
|
1
|
|
|
|
sf
|
1
|
|
1
|
fbod
|
|
fbod
|
1
|
|
1
|
|
|
|
slf
|
1
|
|
1
|
fbod
|
|
fbod
|
1
|
|
1
|
|
|
|
xs
|
|
1
|
1
|
|
fbod
|
fbod
|
|
1
|
1
|
icv-1
|
|
icv-1
|
xbh
|
|
1
|
1
|
|
fbod
|
fbod
|
|
1
|
1
|
icv-1
|
|
icv-1
|
xba
|
|
1
|
1
|
|
fbod
|
fbod
|
|
1
|
1
|
icv-1
|
|
icv-1
|
xpb
|
|
1
|
1
|
|
fbod
|
fbod
|
|
1
|
1
|
icv-1
|
|
icv-1
|
si
|
|
|
|
|
|
|
1
|
|
1
|
|
|
|
xi
|
|
|
|
|
|
|
|
1
|
1
|
icv-1
|
|
icv-1
|
xu
|
|
|
|
|
|
|
|
1
|
1
|
icv-1
|
|
icv-1
|
xgly
|
|
1
|
1
|
|
fbod
|
fbod
|
|
1
|
1
|
icv-1
|
|
icv-1
|
xbt
|
|
1
|
1
|
|
fbod
|
fbod
|
|
1
|
1
|
icv-1
|
|
icv-1
|
xsto
|
|
1
|
1
|
|
fbod
|
fbod
|
|
1
|
1
|
icv-1
|
|
icv-1
|
xii
|
|
|
|
|
|
|
|
|
|
|
1
|
1
|
xmeoh
|
|
|
|
|
|
|
|
|
|
|
1
|
1
|
xmep
|
|
|
|
|
|
|
|
|
|
|
1
|
1
|
xpp
|
|
|
|
|
|
|
|
|
|
|
3
|
3
|
xppr
|
|
|
|
|
|
|
|
|
|
|
3
|
3
|
In terms of variables containing nitrogen, the calculation of the
composite variables is similar to the calculations done in the
Carbon - Nitrogen library. The differences include the addition of
a new state variable, soluble unbiodegradable organic nitrogen
(sni), to the soluble TKN (STKN), the
inclusion of nitrogen components of the new substrate type
(sf), and the inclusion of nitrogen components of
poly‑phosphate-accumulating organisms (xbp). See
Figure 4‑5 for a general diagram of the nitrogen
composite variable calculations.
Figure 4‑5 - Carbon - Nitrogen - Phosphorus
Library: Nitrogen Composite Variables and their Relationships to
the State Variables
Table 4‑7 - CNPLIB
Nitrogen Composite Variables - MANTIS Model
|
STKN
|
XTKN
|
TKN
|
TN
|
sno
|
|
|
|
1
|
snh
|
1
|
|
1
|
1
|
sni
|
1
|
|
1
|
1
|
snd
|
1
|
|
1
|
1
|
xnd
|
|
1
|
1
|
1
|
xbh
|
|
ibhn
|
ibhn
|
ibhn
|
xba
|
|
ibhn
|
ibhn
|
ibhn
|
xi
|
|
iuhn
|
iuhn
|
iuhn
|
xu
|
|
iuhn
|
iuhn
|
iuhn
|
The
calculation for nitrogen state variables differs slightly from
model to model. In particular, the nitrogen fractions of biomass
and other particulate components have different names from model to
model. Table 4‑7 through Table 4‑11
shows the nitrogen composite variable calculations for the
mantis, asm1, asm2d, asm3 and
newgeneral models in CNPLIB.
Table 4‑8 - CNPLIB Nitrogen Composite Variables
- ASM1 Model
|
STKN
|
XTKN
|
TKN
|
TN
|
sno
|
|
|
|
1
|
snh
|
1
|
|
1
|
1
|
sni
|
1
|
|
1
|
1
|
snd
|
1
|
|
1
|
1
|
xnd
|
|
1
|
1
|
1
|
xbh
|
|
ixbn
|
ixbn
|
ixbn
|
xba
|
|
ixbn
|
ixbn
|
ixbn
|
xi
|
|
ixun
|
ixun
|
ixun
|
xu
|
|
ixun
|
ixun
|
ixun
|
Table 4‑9 - CNPLIB Nitrogen Composite Variables
- ASM2d Model
|
STKN
|
XTKN
|
TKN
|
TN
|
sno
|
|
|
|
1
|
snh
|
1
|
|
1
|
1
|
sni
|
1
|
|
1
|
1
|
snd
|
1
|
|
1
|
1
|
xnd
|
|
1
|
1
|
1
|
si
|
insi
|
|
insi
|
insi
|
sf
|
insf
|
|
insf
|
insf
|
xbp
|
|
inbm
|
inbm
|
inbm
|
xbh
|
|
inbm
|
inbm
|
inbm
|
xba
|
|
inbm
|
inbm
|
inbm
|
xi
|
|
inxi
|
inxi
|
inxi
|
xs
|
|
inxs
|
inxs
|
inxs
|
Table 4‑10 - CNPLIB Nitrogen Composite Variables
- ASM3 Model
|
STKN
|
XTKN
|
TKN
|
TN
|
sno
|
|
|
|
1
|
snh
|
1
|
|
1
|
1
|
sni
|
1
|
|
1
|
1
|
snd
|
1
|
|
1
|
1
|
xnd
|
|
1
|
1
|
1
|
xbh
|
|
inbm
|
inbm
|
inbm
|
xba
|
|
inbm
|
inbm
|
inbm
|
xi
|
|
inxi
|
inxi
|
inxi
|
xs
|
|
inxs
|
inxs
|
inxs
|
ss
|
inss
|
|
inss
|
inss
|
si
|
insi
|
|
insi
|
insi
|
Table 4‑11 - CNPLIB
Nitrogen Composite Variables - NEWGENERAL model
|
STKN
|
XTKN
|
TKN
|
TN
|
sno
|
|
|
|
1
|
snh
|
1
|
|
1
|
1
|
sni
|
1
|
|
1
|
1
|
snd
|
1
|
|
1
|
1
|
xnd
|
|
1
|
1
|
1
|
xbp
|
|
fnzh
|
fnzh
|
fnzh
|
xbh
|
|
fnzh
|
fnzh
|
fnzh
|
xba
|
|
fnzh
|
fnzh
|
fnzh
|
xi
|
|
fnzeh
|
fnzeh
|
fnzeh
|
xu
|
|
fnzeh
|
fnzeh
|
fnzeh
|
Generally, soluble total phosphorus (STP) is equal to
the sum of soluble phosphorus (sp) and the phosphorus
components of ss, si, and sf, as appropriate.
Particulate total phosphorus (XTP) is the sum of
stored polyphosphate (xpp and xppr) and the
phosphorus components of xs, xi, xbh, xba, xbp and
xmeoh. See Figure 4‑6 for a schematic of these
calculations. The calculations differ slightly from model to model.
Table 4‑12 to
Table 4‑15 describe the
phosphorus composite variable calculations.
Figure 4‑6 - Carbon - Nitrogen - Phosphorus
Library: Phosphorus Composite Variables and their Relationship to
the State Variables
Table 4‑12 -
CNPLIB Phosphorus Composite Variables - ASM1/MANTIS Models
|
STP
|
XTP
|
TP
|
sp
|
1
|
|
1
|
xi
|
|
ixup
|
ixup
|
xu
|
|
ixup
|
ixup
|
xbh
|
|
ixbp
|
ixbp
|
xba
|
|
ixbp
|
ixbp
|
xbp
|
|
ixbp
|
ixbp
|
xmep
|
|
0.205
|
0.205
|
xpp
|
|
1
|
1
|
xppr
|
|
1
|
1
|
Table 4‑13 – CNPLIB Phosphorus Composite
Variables – ASM3 Model
|
STP
|
XTP
|
TP
|
sp
|
1
|
|
1
|
ss
|
ipss
|
|
ipss
|
si
|
ipsi
|
|
ipsi
|
xi
|
|
ipxi
|
ipxi
|
xs
|
|
ipxs
|
ipxs
|
xbh
|
|
ixbp
|
ixbp
|
xba
|
|
ixbp
|
ixbp
|
xbp
|
|
ixbp
|
ixbp
|
xmep
|
|
0.205
|
0.205
|
xpp
|
|
1
|
1
|
xppr
|
|
1
|
1
|
Table 4‑14 - CNPLIB Phosphorus Composite
Variables - ASM2d Model
|
STP
|
XTP
|
TP
|
sp
|
1
|
|
1
|
sf
|
ipsf
|
|
ipsf
|
si
|
ipsi
|
|
ipsi
|
xi
|
|
ipxi
|
ipxi
|
xs
|
|
ipxs
|
ipxs
|
xbh
|
|
ipbm
|
ipbm
|
xba
|
|
ipbm
|
ipbm
|
xbp
|
|
ipbm
|
ipbm
|
xmep
|
|
0.205
|
0.205
|
xpp
|
|
1
|
1
|
xppr
|
|
1
|
1
|
Table 4‑15 - CNPLIB Phosphorus Composite
Variables - NEWGENERAL Model
|
STP
|
XTP
|
TP
|
sp
|
1
|
|
1
|
xpp
|
|
1
|
1
|
xppr
|
|
1
|
1
|
xi
|
|
fpzeh
|
fpzeh
|
xu
|
|
fpzeh
|
fpzeh
|
xbh
|
|
fpzh
|
fpzh
|
xba
|
|
fpzh
|
fpzh
|
xbp
|
|
fpzh
|
fpzh
|
xmep
|
|
0.205
|
0.205
|
The relationships
between the state variables and composite variables are similar to
the Carbon-Nitrogen-Phosphorus library (CNPLIB). Table
4‑7 through
Table
4‑15 shows the relationships between the state and composite
variables that are used in the models in CNPIPLIB.
The composite
variable calculation schemes in MANTSI2LIB are shown in
Figure 4‑7 to Figure 4‑11. Figure 4‑7 shows the
scheme for estimation of soluble BOD5 (SBOD),
particulate BOD5 (XBOD), BOD5
(BOD), soluble ultimate BOD (SBODU),
particulate ultimate BOD (SBODU), ultimate BOD
(BODU), soluble COD (SCOD), particulate
COD (XCOD) and COD (COD). The list of stoichiometric
parameters used in the estimation of composite variables is
provided in Table 4‑16.
Table 4‑16 -
Stoichiometry Parameters used in Estimation of Composite
Variables
Stoichiometry Parameter
|
Default Value
|
Description
|
yhglobal
|
0.666
|
Heterotrophic
biomass yield
|
fuu
|
0.206
|
Fraction of
unbiodegradable residue in biomass
|
fssbodtosscod
|
0.717
|
BOD5
to COD ratio of soluble substrate
|
fpsbodtopscod
|
0.58
|
BOD5
to COD ratio of particulate substrate
|
fbbodtobcod
|
0.566
|
BOD5
to COD ratio of biomass
|
icodtovssxbt
|
1.674
|
xCOD/VSS ratio
of PHA
|
The default value of the stoichiometry parameters can be changed in
the
System > Input Parameters > Global Fixed
stoichiometry menu.
Figure 4‑8 shows the scheme for estimation of
Volatile Suspended Solid (VSS), Inorganic Suspended Solid
(XISS) and Total Suspended Solid (TSS) concentration.
In the estimation of VSS, each biomass concentration is
multiplied by a corresponding VSS to COD factor. For
example, ivsstocodxbh is the VSS to COD
ratio for heterotrophic biomass, xbh. The VSS
to COD ratio for each biomass type are calculated based on
the biomass composition. The default composition of biomass can be
accessed and changed in the
System > Input Parameters > Global Fixed
stoichiometry menu.
As the composition of xs and xi are not well known,
the VSS to COD ratios for these states
ivsstocodxs and ivsstocodxi are
provided as direct inputs. These ratios can be accessed and changed
in System > Input Parameters > Influent
Stoichiometry menu. The state variable of xns is
included in the calculation of VSS for being consistent with
the practice of considering the N fraction in biomass as part of
VSS. A stoichiometry factor of 17.0/14.0 is used to convert
N to NH3.
In the calculation
of Inorganic Suspended Solid (XISS), each biomass
concentration is multiplied by the inorganic fraction in the
biomass. The inorganic fraction for each biomass is estimated by
subtracting the VSS to SS ratio for each individual biomass from
one. The VSS to SS ratio for each individual biomass
type is calculated by using the set biomass composition, for
example ivsstossxbh is the VSS to SS
ratio calculated for the xbh biomass type.
The composite
variable of XISS also includes the contribution from the
inorganic states in the model. The mass contributions from the
xpp and xps states are calculated by using a
stoichiometry factor of 95/31. The factor reflects the conversion
from molecular weight of P to molecular weight of
PO43-. For all the other inorganic states a
stoichiometry ratio of 1 is used.
The Total Suspended
Solid (TSS) concentration is calculated by the sum of
estimated VSS and XISS.
Figure
4‑9 shows the scheme for estimation of soluble part of
Total Kjeldahl Nitrogen (STKN), particulate part of Total
Kjeldahl Nitrogen (XTKN), Total Kjeldahl Nitrogen
(TKN), Total Nitrogen (TN) and Total Nitrogen
including dissolved Nitrogen (TN & dissolved gas). In
the estimation of STKN, the stoichiometry ratio of,
insi is used to estimate the organic nitrogen present in the
si state variable. In the estimation of XTKN,
each biomass concentration is multiplied by a corresponding
stoichiometry factor representing the N content in the
corresponding biomass. For example, inxbh is the
stoichiometry factor representing the N content in the
heterotrophic biomass, xbh. The N fraction for each
biomass type is calculated based on the biomass composition. The
default composition of biomass can be accessed and changed in the
System > Input Parameters > Global Fixed
stoichiometry menu. In the calculation of XTKN, the
N contained in the MgNH4PO4 is also included.
Although, in strict sense this is not a part of the organic
nitrogen, it is assumed that the ammonia contained in the
precipitate shall reflect in the analytical measurement of
TKN.
Figure
4‑10 shows the scheme for estimation of soluble part of
Total Phosphorus (STP), particulate organic part of Total
Phosphorus (XTOP), particulate inorganic part of Total
Phosphorus (XTIP) and Total Phosphorus. In the estimation of
STP, the stoichiometry ratio of, ipnsi is used to
estimate the phosphorus present in the si state
variable. In the estimation of XTOP, each biomass
concentration is multiplied by a corresponding stoichiometry factor
representing the P content in the corresponding biomass. For
example, ipxbh is the stoichiometry factor
representing the P content in the heterotrophic biomass,
xbh. The P fraction for each biomass type is
calculated based on the biomass composition. The default
composition of biomass can be accessed and changed in the
System > Input Parameters > Global Fixed
stoichiometry menu. In the calculation of XTIP, the
P contained in various P- precipitates is included. The
stoichiometry factor for each precipitate are estimated based on
the composition of the precipitate. These stoichiometry factors are
also available in Global Fixed stoichiometry menu. The
composite variable of XTP is estimated by the sum of
XTOP and XTIP. The TP is estimated by the sum of
STP and XTP.
Figure
4‑11 shows the scheme for estimation of soluble part of
Total Organic Carbon (STOC), particulate part of Total
Organic Carbon (XTOC), and Total Organic Carbon. In the
estimation of STOC, the stoichiometry ratio of icsac,
icsmet, icspro are estimated based on the substrate
compositions. The stoichiometry factor of icscol,
icsi and icss on the other hand needs to be set by
direct user input. These factors can be accessed and changed
in
System > Input Parameters > Influent
stoichiometry menu. In the estimation of XTOC, each
biomass concentration is multiplied by a corresponding
stoichiometry factor representing the C content in the
corresponding biomass. For example, icpxbh is the
stoichiometry factor representing the C content in the
heterotrophic biomass, xbh. The C fraction for each
biomass type is calculated based on the biomass composition. The
default composition of biomass can be accessed and changed in the
System > Input Parameters > Global Fixed
stoichiometry menu. For the particulate states of xi
and xs, the user can directly enter the C content in
the
System > Input Parameters > Influent
stoichiometry menu. The composite variable of TOC is
estimated by the sum of XTOC and STOC.
Table
4‑17 presents the summary of the stoichiometry parameters
used in the MANTSI2LIB and access menus for changing the default
values.
Table 4‑17 -
Access Menus for Different Stoichiometry Parameters in
MANTIS2LIB
Stoichiometry Parameter
|
Access Menu
|
yhglobal
|
System > Input Parameters > Global
Fixed stoichiometry
|
fuu
|
System > Input Parameters > Global
Fixed stoichiometry
|
fssbodtosscod
|
System > Input Parameters > Global
Fixed stoichiometry
|
fpsbodtopscod
|
System > Input Parameters > Global
Fixed stoichiometry
|
fbbodtobcod
|
System > Input Parameters > Global
Fixed stoichiometry
|
COD to VSS ratio for biomass
|
System > Input Parameters > Global
Fixed stoichiometry
(Calculated based on biomass
composition)
|
COD to VSS ratio for xs and
xi
|
System > Input Parameters >
Influent stoichiometry
|
VSS to SS ratio for biomass
|
System > Input Parameters > Global
Fixed stoichiometry
(Calculated based on biomass
composition)
|
N fraction in biomass
|
System > Input Parameters > Global
Fixed stoichiometry
(Calculated based on biomass
composition)
|
N fraction in xi, si
|
Influent Characterization menu
of influent object
|
P fraction in biomass
|
System > Input Parameters > Global
Fixed stoichiometry
(Calculated based on biomass
composition)
|
P fraction in xi, si
|
Influent Characterization menu
of influent object
|
C fraction in biomass and sac,
spro, smet
|
System > Input Parameters > Global
Fixed stoichiometry
(Calculated based on biomass/substrate
composition)
|
C fraction in xcol,xss,si,xi,xs
|
System > Input Parameters >
Influent stoichiometry
|
Figure 4‑7 – MANTIS2LIB – Calculation Procedure
for Composite Variables SCOD, COD, SBOD, BOD, SBODU and
BODU
Figure 4‑8 - MANTIS2LIB - Calculation Procedure
for Composite Variables VSS, TSS
Figure 4‑9 - MANTIS2LIB - Calculation Procedure
for Composite Variables STKN and TKN
Figure 4‑10 - MANTIS2LIB - Calculation
Procedure for Composite Variables STP, XTP, and TP
Figure 4‑11 - MANTIS2LIB - Calculation
Procedure for Composite Variables STOC and TOC
Displaying Composite
Variables
All of the
composite variables, including the most common ones such as X
(total suspended solids), TKN (Total Kjeldahl Nitrogen), BOD, COD,
etc., can be accessed by selecting the Output Variables >
Concentrations menu. These variables (and those found by
clicking on the More... button) can be tagged and
placed on output graphs.
Whether
stoichiometry has been set globally or locally within each object,
the list of composite variables that are available to be displayed
will depend on which library has been used for that particular
GPS-X layout (CNLIB, CNPLIB, etc.).
It is possible to
use a model in the layout that does not simulate the fate of all
components of a given library (e.g. mantis in CNPLIB, which
does not model phosphorus even though there are phosphorus state
variables in CNPLIB). In these cases, these extra components will
be modelled as if they were inert (i.e. no biological
transformation applied, but mixing and settling are applied).
The composite
variable list displayed in the Output Variables >
Concentrations menu is the list associated with the library
used to create the layout (e.g. CNPLIB). All composite variables
are calculated for the given library, even if the component states
are not modelled biologically in the reactor. For the example above
(mantis model in CNPLIB), soluble phosphorus (sp) is
not modelled biologically, and therefore behaves the same as
si (soluble inerts); however, because CNPLIB contains the
composite variable TP (total phosphorus), it will be calculated
(from sp and other components) even though sp was
modelled as an inert.
It is important to
take this into consideration when selecting which model and library
you will be using when creating your plant layout.
CHAPTER
5
For every influent
model that is used in a wastewater plant layout, it is important to
properly specify the influent characteristics and the influent
stoichiometry.
To help users
better understand influent characterization, a special utility
program, called Influent Advisor, was developed by
Hydromantis. The tool helps users to visualize and debug
influent characterization data. It is recommended that users
make use of this utility tool so that influent characterization
errors can be avoided.
The mathematical
description of the influent wastewater that is fed to the plant
model is the single most important aspect of a simulated system.
Without significant consideration of the influent characterization,
the plant model will be limited in its ability to predict the
dynamic behavior of the plant.
To access the
Influent Advisor tool, right-click on any wastewater
influent object, and select the Composition > Influent
Characterization menu item.
Figure 5‑1 -
Opening the Influent Advisor Tool
The Influent
Advisor screen shows three columns of data: User
Inputs, State Variables, and Composite Variables, as
shown in Figure 5‑2.
Figure 5‑2 - Influent Advisor Menu
The fields in the
left-hand column show the inputs available to the user, such as
influent concentrations and stoichiometric ratios. The centre
and right hand columns show the state variable and composite
variable concentrations calculated from the user inputs.
As the user changes
values in the left-hand column, the variables in the centre and
right-hand columns are automatically updated. This allows for
easy debugging of confusing or conflicting influent
characterization data.
Clicking on any
variable in the centre and right-hand column illustrates how that
value is calculated. The formula will be displayed in the formula
box (located in the lower middle of the screen – you may need to
scroll down to see it). The values used in the formula will
be highlighted in the tables so that the applicable cells can be
identified. If a negative value is calculated in either table, the
cell will turn red. Correcting problematic data is only a matter of
adjusting the input cells to achieve non-negative values.
Figure 5‑3 -
Influent Advisor Screen Showing Highlighted Cells and Negative
Values
(Highlighted in Red)
The influent models
used in GPS-X make certain assumptions about what data may or may
not be available. For instance, no influent model uses both COD and
BOD data, even though this data may be available and may provide
important information about how that organic material is
partitioned into the available state variables. Influent Advisor
helps overcome this shortcoming.
Take an example in
which BOD and COD data is available and the BODbased model is
chosen for the influent object. The applicable data can be entered
into Influent Advisor, including the available BOD data. The user
can then scroll to the right table and check the COD value
calculated based on the input data. If the COD data is in agreement
with the measured COD data (assuming no negative values in any
cells), then the user can be assured that the input data is
consistent with the available data. If the COD data is different
from the measured COD, then the unknown (or estimated) input data
should be adjusted until acceptable agreement is achieved.
NOTE:
Correctly setting up the influent is critically important to the
simulation; therefore, a set of warning messages has been
developed, and will appear in the
Command window, when necessary. For instance, if
the user inadvertently enters a value for xsto, but has
chosen ASM1 as the local biological model, then a warning
message (`time = <timestamp>
xsto<streamlabel> is non-zero, xsto is not a state variable
in ASM1') will appear in the
Log window. This is a signal to go back to the
influent data forms and correct a problem with the influent. It is
recommended that
Influent Advisor be used to debug your influent
characterization. Hydromantis recommends that users make full
use of the Influent Advisor as a tool to help identify problems,
and understand the interconnectivity between state and composite
variables.
During a
simulation, error messages related to the influent may appear in
the simulation Log window. These messages will most likely
be the result of improper influent stoichiometry. If an error
message does appear, then the influent stoichiometry should be
examined for possible errors.
There are five (5)
influent objects in GPS-X:
Table 5‑1 – Influent Objects
Name
|
Object
|
Use
|
Models Available
|
Wastewater
Influent
|
|
Continuous wastewater flows (steady or dynamic)
|
bodbased
codbased
codstates
sludge
states
tssfrac
|
Batch Influent
|
|
Batch deliveries of septage or other discontinuous wastewater
flows
|
bodbased
codbased
codstates
sludge
states
tssfrac
|
Water Influent
|
|
Clean water input (steady or dynamic)
|
water
|
COD Chemical
Dosage
|
|
Dosage of COD into streams or objects
|
codfeed
|
Acid Dosage
|
|
Dosage of acid for pH control
|
acidfeed
|
The influent
objects contain models, options, and features that are relevant to
the type of influent being used. For example, the continuous
wastewater model has options for specifying a diurnal pattern for
influent flow, a feature not found in the chemical dosage
object.
Wastewater Influent
Object
The wastewater
influent object (brown arrow) is used to characterize continuous
streams of wastewater flow. The continuous influent flow rate
is specified in the influent object’s Flow > Flow
Data menu. (Figure 5‑4 and Figure
5‑5)
Figure 5‑4 - Selecting the Influent Flow Data
Menu
Figure 5‑5 - Influent Flow Data Menu, showing
Flow Type Options
The flow is
specified via one of the four methods:
1.
Data – users set the flow rate directly, via menu entry or
read from file.
2.
Sinusoidal – GPS-X applies a sinusoidal curve to the
influent flow set in the menu or read from file.
3.
Diurnal Flow – a daily diurnal patter is set via flow rates
at different times of the day.
4.
Diurnal Flow Factor – a daily diurnal pattern is set via
flow rates at different times of the day.
The influent models
available in GPS-X depend on the model library and the local
biological model used to relate the state variables to the
composite variables. The manner in which the state variables are
calculated sometimes differs from model to model and library to
library. This chapter discusses the models available and the flow
choices available with the batch influent object.
Figure 5‑6 -
The Influent Models
Each influent model
calculates a complete set of library-dependent state variables that
are passed to the rest of the plant layout. The influent models
differ only in the type of information required as inputs to the
model.
The descriptions of
the influent models that follow are an overview of how they work;
however, due to their complexity and their dependence on the local
biological model, and the library currently in use, users are
referred to the Influent Advisor to understand the
calculations being made in each model.
BODbased
The BODbased
influent model is the choice when BOD data is available and COD
data is not available; however, due to the approximations and the
nature of the BOD measurement, special care must be taken to
properly estimate influent particulate inerts. If this model is
selected, the user inputs total carbonaceous BOD5, total TKN, total
suspended solids, a few state variables. The state variables are
normally zero, except for the soluble inert organics, soluble
ortho-phosphate (CNPLIB) and alkalinity (Figure 5‑7
and Figure 5‑8 and several stoichiometric
fractions).
These inputs are used to calculate the remaining influent state
variables: readily biodegradable substrate (ss), slowly
biodegradable substrate (xs), particulate inert organics (xi), free
and ionized ammonia (snh), particulate biodegradable organic
nitrogen (xnd), and soluble biodegradable organic nitrogen
(snd).
Figure 5‑7 – MANTIS2 Library BODbased Influent
Model Inputs
The relationships between the calculated state variables and the
composite and stoichiometric fractions are library-specific and the
relationships are shown in the appendix to this chapter
(Figure 5‑23 and Figure 5‑24).
BODultimate is assumed to be equivalent to the biodegradable
COD.
The stoichiometric fraction fss (soluble
substrate:BODultimate), is used to determine what fraction of the
total BODultimate is soluble substrate (sf for asm2d
orss for all other biological models). In terms of sampling
measurements, the fraction fss can be estimated from
the ratio of filtered BOD20
:BOD20. The particulate substrate state variable
(xs) is then calculated by difference (i.e. in cnlib,
xs = BODultimate * (1-fss) -xba-xbh-xsto)). The
particulate inert organics (xi) state variable is calculated
by subtracting all the other particulate carbonaceous organic
states (xsto, xbh, xba, xu (all input by user) and xs
(calculated above)) from the particulate COD
(XCOD).
Calculation of the
three unknown nitrogen state variables depends on the calculation
of TKN which is biological model-dependent. For instance,
with the asm1 or mantis models the amount of free and
ionized ammonia (snh) is the fraction of ammonia
(fnh)multiplied by the TKN. The remainder is
the biodegradable organic nitrogen, which is partitioned into three
areas including nitrogen associated with particulate state
variables (i.e. xbh, xba, xi and xu), and soluble and
particulate organic nitrogen (snd and xnd) using the
fxn fraction and the nitrogen fractions of the
particulate state variables (Figure 5‑8).
Figure 5‑8 – MANTIS2 Library Nutrient Fractions
for the BODbased Influent Model
This differs from
the calculations made when asm3 has been chosen as the local
biological model because snd and xnd are not state
variables in asm3. In this case, soluble TKN is calculated
as the difference between the total TKN and the particulate TKN.
The nitrogen associated with the soluble state variables (ss
and si) is subtracted from the soluble TKN and the remaining
TKN is equated to ammonia (snh). Hence, in the asm3
BOD based influent model, fnh and fxn are not used.
Influent Advisor will help users determine what parameters
are necessary for each mode.
CODFractions
This model requires
an input of total COD, total TKN, total phosphorus (in
CNPlib), state variables (the state variables are normally
zero except for ammonia, soluble ortho‑phosphate and alkalinity)
and several stoichiometric fractions. From these inputs, the
complete set of state variables, composite variables and nutrient
fractions are calculated (Figure 5‑18 and
Figure 5‑19).
The calculation of
the nitrogen and phosphorus (CNPlib) state variables and
fractions is complicated in the codfractions influent
models. This is because some of the nitrogen and phosphorus in the
influent is associated with the organic state
variables.(i.e. N content of active biomass). Therefore,
the model must adjust itself depending on the applicable composite
variable model and the current organic states. As with the other
influent models, these nutrient fractions are specified in this
model. However, in the event that a mass balance is not achievable
with the user input data, the codfractions model will
recalculate these nutrient fractions to force the mass balance. It
is recommended that Influent Advisor be used to set-up and
understand these calculations.
This model has the
advantage that each of the calculations is based on the total COD,
total TKN, total phosphorus (CNPlib), soluble
ortho-phosphate (CNPlib) and ammonia inputs. Therefore, (as
often is available in practice) a series of these data over a
period of time can be read in from a data file and the influent
state variables will vary with the load. This is in contrast to
other influent models in which some state variables are input
directly and will not automatically vary with a load change.
CODStates
This model works
similarly to the codfractions influent model, however all COD input
fractions are set as a fraction of total COD. This allows
users to specify total COD, TKN and ammonia as their main
characterization inputs. Soluble inert COD (si),
readily biodegradable substrate (ss, sf or slf),
particulate inert material (xi), unbiodegradable cell
products (xu) and biomass concentrations (xbh, xba
and xbp) are specified via fractions of total COD, as shown
in Figure 5‑9.
Figure 5‑9 - CODstates Influent Model
Inputs
This model has been
designed to mimic the input of a sludge stream. The user inputs the
total suspended solids, a few state variables, and a couple of
stoichiometric fractions. From these inputs, the organic solids are
partitioned into heterotrophic biomass (i.e. degradable particulate
material), polyphosphate accumulating biomass (CNPLIB only)
and un-biodegradable particulate material. The remaining
particulate organic state variables are set to zero. All the
soluble state variables are defaulted to zero in this model, except
for dissolved oxygen and alkalinity.
This model is
appropriate if a full influent characterization has been performed
and the influent state variables have been calculated manually. If
the user selects the states model, input values for the
state variables, and a few stoichiometric fractions used for
calculating the composite variables are required. The CNlib states
data entry form (Composition > Influent
Composition) is shown in Figure 5‑9, and the
stoichiometry entry form (Composition > Influent
Stoichiometry) is shown in Figure 5‑10.
Figure 5‑10 - CN Library States Influent Model
Influent Stoichiometry Inputs
TSSCOD
The tsscod
influent model can be used successfully if the influent was
characterized using COD and suspended solids as the main
components. The tsscod influent model was developed based on the
Activated Sludge Model No. 2 report (Henze et
al., 1995).
If this model is
selected, the user inputs total COD, total TKN, total suspended
solids, a few state variables (the state variables are normally
zero except for the soluble inert organics, soluble ortho-phosphate
(CNPLIB) and alkalinity and several stoichiometric
fractions. These inputs are then used to calculate the remaining
state variables (Figure 5‑7 and Figure 5‑8).
Particulate COD
(XCOD) is calculated from the TSS using two
stoichiometric fractions. This is then divided into its component
parts via stoichiometric fractions or explicitly as read from the
data input forms leaving the particulate inert organics component
(xi) to be calculated by subtraction.
Figure 5‑12 shows how the soluble COD
(SCOD) components are calculated from the COD, XCOD
and the stoichiometric parameter, frsi. See the
bodbased model for the calculation of the nitrogen and
phosphorus (CNPLIB) state variables. It is recommended that
Influent Advisor be used to understand these
calculations.
Figure 5‑11
- CN Library tsscod Model Particulate Inert Calculation
Figure 5‑12 - CN Library tsscod Influent
Soluble Components Calculation
Runoff Model
The runoff
flow model uses a parallel linear reservoir model to simulate wet
weather flow in sanitary and combined sewer systems. This model is
not a mechanistic hydrological model, but a simple mathematical
transformation.
The equations
are:
Equation 5.1
and
Equation 5.2
where:
Pd =
rainfall that enters the sewer system directly
Pi =
rainfall that enters the sewer system indirectly
Ptotal = total rainfall
over the catchment area
Cd = fraction
of total rainfall that enters the sewer system directly
Ci =
fraction of total rainfall that enters the sewer system
indirectly
Total runoff (Qtotal) is calculated with the following
equation:
Equation 5.3
which is based in the following equations:
Equation 5.4
Equation 5.5
Equation 5.6
Equation 5.7
where:
Kd =
decay rate of linear reservoir representing direct runoff
Ki
= decay rate of linear reservoir representing indirect
runoff
A =
total catchment area
The chemical dosage
influent object (blue arrow) is used to characterize streams which
are not typical wastewater inputs, but rather chemical or water
inputs to a wastewater treatment process. The models found in
the chemical dosage influent have been developed specifically for
this purpose, and are set up for easy conversion of typical
chemical components into the state variables used in the biological
models. The flow is specified only through the Data
option as described in Figure 5‑5.
Acetate
The acetate
influent model can be used to simulation the addition of acetate to
the treatment process. The acetate dose (as acetic acid) can be
entered as a percentage of purity, or in a variety of units (mol
acetate/L, g (acetate*COD)/m3, g(acetate)/L, etc.), by
selecting from the drop-down units menu (Figure
5‑13). The COD equivalent of the acetate dose is
automatically converted to slf if asm2d or
newgeneral have been chosen for the local biological model
or ss (readily biodegradable substrate) for all other local
biological models.
Figure 5‑13 - CN Library Acetate Influent Model
- Acetate Dose Form
The methanol influent model can be used to simulate the
addition of methanol to the treatment process. The methanol dose
can be entered as a percentage of purity, or in a variety of other
units (mol methanol/L, g(methanol*COD)/m3,
g(methanol)/L, etc.), by selecting from the units drop-down menu
(Figure 5‑14). The COD equivalent of the methanol
dose is automatically converted to sf if asm2d has
been chosen for the local biological model or ss (readily
biodegradable substrate) for all other local biological models.
Figure 5‑14 - CN Library Methanol Influent
Model Inputs
This influent model
was developed to simulate a rain event or alkalinity addition
whereby the influent hydraulic load could be increased without
increasing the organic or nitrogen load to the plant. In the water
influent, the only variable to be provided in the composition menu
is the alkalinity. All other state variables are set to zero.
The difference
between the batch influent (truck) object and the continuous
influent objects (arrows) concerns the flow and load types. The
available flow and load types for the batch influent object are
averageand individual.
In the batch influent, if the
average flow and load types are selected, the model will
behave exactly like the continuous influent.
If the
individual flow type is selected, there
will be an intermittent (or batch) influent. Under the Flow
sub-menu item individual, the user can specify
the starting and ending time of the batch influent, and the volume
of each truck (1 truck per day by default - the number of trucks
per day is specified in the Influent Composition
sub-menu). The total volume specified will be fed at an
average rate over the total dumping time specified, that is there
will be one influent flow spike.
If the individual load
type is selected, the loading is determined by the amount specified
in the Individual Loadssub-menu. The stoichiometry is
specified in the Influent Characterizationmenu and the
number of trucks per day is specified in the
Compositionsub-menu (see Figure
5‑15).
Figure 5‑15 - Batch Input Menu - Flow Data
Model Inputs
Model-Dependent State and
Composite Variables
When setting up
your influent model, it is important that you realize what state
variables are used in each biological model. The following sections
outline the state variables used in the models in the different
libraries.
Of similar
importance is an understanding of how the composite variables are
calculated from the state variables. This is important because many
of the influent models go backwards (i.e. from the composite
variables to the state variables); hence, an understanding of these
relationships will help if debugging is necessary. Influent
Advisor has been developed specifically to help navigate the
more complicated influent models. The following figures were
developed to help with this understanding.
The composite variable figures
are copies of the figures presented in the composite variables
chapter of this Reference; however, they are repeated here because
of their importance to the influent model calculations.
CN and CNIP Libraries
This section
contains one table and two figures which are to help the user
understand how GPS-X has calculated the CNLIB state variables, and
what state variables should be calculated depending on which
biological model you are using.
Table 5‑2 -
State Variables Used in Each Biological Model Included in CNLIB and
CNIPLIB
State Variable
|
Mantis
|
ASM1
|
ASM3
|
si
|
✓
|
✓
|
✓
|
ss
|
✓
|
✓
|
✓
|
xi
|
✓
|
✓
|
✓
|
xs
|
✓
|
✓
|
✓
|
xu
|
✓
|
✓
|
|
xsto
|
|
|
✓
|
xbh
|
✓
|
✓
|
✓
|
xba
|
✓
|
✓
|
✓
|
so
|
✓
|
✓
|
✓
|
snh
|
✓
|
✓
|
✓
|
snd
|
✓
|
✓
|
|
xnd
|
✓
|
✓
|
|
sno
|
✓
|
✓
|
✓
|
snn
|
✓
|
✓
|
✓
|
salk
|
✓
|
✓
|
✓
|
xii
|
✓
|
✓
|
✓
|
Figure 5‑16
– CN Library Organic State and Composite Variables
Figure 5‑17
- CN Library Nitrogen State and Composite Variables
CNP
and CNPIP Libraries
This section
contains one table and three figures which are to help the user
understand how GPS-X has calculated the CNPLIB state variables, and
what state variables should be calculated depending on which
biological model you are using.
Table 5‑3 – State Variables Used in Each
Biological Model included in CNPLIB and CNPIPLIB
State Variable
|
Mantis
|
ASM1
|
ASM2d
|
ASM3
|
NewGeneral
|
si
|
✓
|
✓
|
✓
|
✓
|
✓
|
ss
|
✓
|
✓
|
|
✓
|
✓
|
sf
|
|
|
✓
|
|
|
slf
|
|
|
✓
|
|
✓
|
xi
|
✓
|
✓
|
✓
|
✓
|
✓
|
xs
|
✓
|
✓
|
✓
|
✓
|
✓
|
xu
|
✓
|
✓
|
|
|
✓
|
xbh
|
✓
|
✓
|
✓
|
✓
|
✓
|
xba
|
✓
|
✓
|
✓
|
✓
|
✓
|
xbp
|
|
|
✓
|
|
✓
|
xbt
|
|
|
✓
|
|
✓
|
xsto
|
|
|
|
✓
|
|
so
|
✓
|
✓
|
✓
|
✓
|
✓
|
sp
|
|
|
✓
|
|
✓
|
xpp
|
|
|
✓
|
|
✓
|
xppr
|
|
|
|
|
✓
|
snh
|
✓
|
✓
|
✓
|
✓
|
✓
|
snd
|
✓
|
✓
|
|
|
✓
|
xnd
|
✓
|
✓
|
|
|
✓
|
sno
|
✓
|
✓
|
✓
|
✓
|
✓
|
snn
|
✓
|
✓
|
|
✓
|
|
salk
|
✓
|
✓
|
✓
|
✓
|
✓
|
xmeoh
|
|
|
✓
|
|
|
xmep
|
|
|
✓
|
|
|
xii
|
✓
|
✓
|
✓
|
✓
|
✓
|
Figure 5‑18 - CNP Library Organic State and
Composite Variables
Figure 5‑19 - CNP Library Nitrogen State and
Composite Variables
Figure 5‑20
- CNP Library Phosphorus State and Composite Variables
There are seven influent objects in MANTIS2LIB of GPS-X:
Table 5‑4 - Influent Objects in
MANTIS2LIB
Name
|
Object
|
Use
|
Models Available
|
Wastewater
Influent
|
|
Continuous wastewater flows (steady or dynamic)
|
bodbased
codbased
codstates
sludge
states
tssfrac
|
Batch Influent
|
|
Batch deliveries of septage or other discontinuous wastewater
flows
|
bodbased
codbased
codstates
sludge
states
tssfrac
|
Water Influent
|
|
Clean water input (steady or dynamic)
|
water
|
Stormwater Runoff
|
|
Clean water input from storm events
|
runoff
|
COD Chemical
Dosage
|
|
Dosage of COD into streams or objects
|
codfeed
|
Acid Dosage
|
|
Acid addition for pH control
|
acidfeed
|
Alkali Dosage
|
|
Alkali addition for pH control
|
alkalifeed
|
Nutrient Dosage
|
|
Nutrient addition
|
nutrifeed
|
The eight influent objects contain
models, options and features that are relevant to the type of
influent being used. For example the continuous wastewater
model has options for specifying a diurnal pattern for influent
flow, a feature not found in the other influent objects. The
influent models in MANTIS2LIB are similar to influent models in
other libraries with a few differences.
Wastewater Influent
Object
The flow rate setup
in wastewater influent object is similar to the flow rate setup in
other libraries. The built-in influent advisor also works in a
similar way as for the other libraries. However, a few key
differences are with respect to how the stoichiometric parameters
for composite variable calculations are organized and calculated in
this library. In MANTIS2LIB, the stoichiometric parameters are
accessed through the
System > Input Parameters > Biochemical Model
Settings menu (Figure 5‑21). The
stoichiometric parameters available in the COD to VSS ratio
and Fractions Used in Composite Variable Calculations are
shown in Figure 5‑22 and Figure 5‑23.
The stoichiometric parameters available in the COD to VSS
ratio group are the parameters which are influent specific and
representative of a composite component of unknown composition. On
the other hand, the stoichiometry parameters available in the
Fractions Used in Composite Variable Calculations are
parameters which can be calculated based on the chemical
composition (acetic acid, methanol etc.) of the component or some
underlying fundamental estimation procedure. The default values of
inorganic fractions (N, P, etc.) in different type of biomass may
be accessed by pressing the More… button on the Fractions
Used in Composite Variable Calculations group (Figure
5‑24).
Figure 5‑21 - Accessing the Stoichiometry
Parameters in MANTIS2LIB
Figure 5‑22 - Influent Specific Stoichiometric
Parameters
Figure 5‑23 - Fixed Stoichiometric Parameters
in MANTIS2LIB
Figure 5‑24 - More... Fixed Stoichiometric
Parameters in MANTIS2LIB
Batch Influent Object
The batch influent
object in MANTIS2LIB is similar to the batch influent object in
other libraries.
Water Influent Object
The water influent
object is similar to the batch influent object in other
libraries.
Chemical Dosage
Object
The COD chemical dosage object uses the codfeed model
(Figure 5‑25). The COD chemical dosage object allows user to select
the type of COD used in the feed. Six COD sources acetic acid,
propionic acid, methanol, molasses, glycerol and generic mixed
substrate are available for selection from the Composition >
Feed Chemical Details menu (Figure 5-26). The Feed Chemical
Details menu is as shown in Figure 5‑27.
For pure organic chemical acetic acid, propionic acid, methanol and
glycerol, two input parameters of % purity and density of chemical
solution at the selected %-purity are required as input. Depending
on the chemical selection, the concentration of the corresponding
state variable is set. For example, if acetic acid is selected, the
concentration of acetate (sac) is set to the value determined by
the set % purity and density of the chemical solution. For molasses
and mixed substrate, in addition to the density and % purity of the
compound, N/COD and P/COD ratios in the substrate may be specified
by the user. The nitrogen compound in the substrate is assigned to
the soluble organic nitrogen state variable (snd) while the
phosphorus in chemical is assigned to the soluble ortho-P state
variable (sp).
Figure 5‑25 -
Models in COD Chemical Dosage Influent Object
Figure 5‑26 - Accessing Feed Chemical Details
Menu
Figure 5‑27 - Selection of Feed Chemical and
Set-up of Chemical Properties
The COD chemical
dosage object has a built-in flow rate controller. The controller
is useful for controlling the feed rate of COD based on a user
defined controlled variable.
Acid Dosage Object
The acid dosage
object is only available in the MANTIS2LIB. The acid dosage object
uses the acidfeed model (Figure 5‑28). The
acid dosage object allows user to select the type of acid used in
the feed. Three acids HCl, H2SO4 and
HNO3 are available for selection from the Composition
> Feed Chemical Details menu (Figure 5‑29).
The Feed Chemical Details menu is as shown in Figure
5‑30. For the selected chemical, two input parameters of %
purity and density of chemical solution at the selected %-purity
are required. If HCl or H2SO4 is selected,
the state variable of other anion (sana) is set to the
equivalent dosed amount. If HNO3 is the selected acid
then, the state variable of Nitrate-N (snoa) is set to an
equivalent concentration.
Figure 5‑28 - Models in Acid Dosage Influent
Object
Figure 5‑29 - Accessing Feed Chemical Details
Menu
Figure 5‑30 - Selection of Feed Chemical Set-up
of Chemical Properties
The acid dosage
object has a built-in flow rate controller. The controller is
useful for controlling the feed rate of acid based on a control
variable (i.e. pH) in a reactor of interest.
Alkali Dosage Object
The alkali dosage
object is only available in the MANTIS2LIB. The alkali dosage
object uses the alkalifeed model (Figure
5‑31). The alkali dosage object allows user to select the
type of alkali (base) used in the feed. Six alkalis NaOH,
Ca(OH)2, Mg(OH)2, NaHCO3,
CaCO3 and Na2CO3 are available for
selection from the Composition > Feed Chemical Details
menu (Figure 5‑32). The Feed Chemical Details
menu is as shown in Figure 5‑33. For the selected
chemical, two input parameters of % purity and density of chemical
solution at the selected %-purity are required. The chemical and
corresponding state variables which are set in the feed are shown
in Table 5‑5.
Figure 5‑31 - Models in Alkali Dosage Influent
Object
Figure 5‑32 - Accessing Feed Chemical Details
Menu
Figure 5‑33 - Selection of Feed Chemical and
Setup of Chemical Properties
Table 5‑5 -
Alkali Chemicals and Affected States in the Feed
Chemical
|
Affected States in Feed
|
NaOH,
|
sana
|
Ca(OH)2,
|
sca
|
Mg(OH)2,
|
smg
|
NaHCO3
|
sana, stic
|
CaCO3
|
sca, stic
|
Na2CO3
|
sana, stic
|
The alkali dosage object has a built-in flow rate controller. The
controller is useful for controlling the feed rate of alkali based
on a control variable (e.g. pH) in a reactor of interest.
Nutrient Dosage
Object
The nutrient dosage
object is only available in the MANTIS2LIB. The nutrient dosage
object uses the
nutrifeed model (Figure 5‑34). The
nutrient dosage object allows user to select the type of nutrient
used in the feed. Four nutrients NH4Cl, Urea,
(NH4)3PO4 and
H3PO4 are available for selection from the
Composition > Feed Chemical Details menu
(Figure 5‑35). The Feed Chemical
Details menu is as shown in Figure 5‑36. For
the selected chemical, two input parameters of % purity and density
of chemical solution at the selected %-purity are required. The
chemical and corresponding state variables which are set in the
feed are shown in
Figure 5‑34 - Models in Nutrient Dosage
Influent Object
Figure 5‑35 - Accessing Feed Chemical Details
Menu
Figure 5‑36 - Selection of Feed Chemical and
Setup of Chemical Properties
Table 5‑6 - Nutrient Chemicals and
Affected States in the Feed
Chemical
|
Affected States in Feed
|
NH4Cl
|
snh, sana
|
Urea
|
snd
|
(NH4)3PO4
|
snh, sp
|
H3PO4
|
sp
|
MgCl2
|
smg, sana
|
The nutrient dosage object has a built in flow rate controller. The
controller is useful for controlling the feed rate of nutrient for
a user-defined control variable.
CHAPTER 6
This chapter
examines the suspended-growth models that are available in the
different GPS-X libraries. Each of the biological models
available in GPS-X is implemented in both the completely mixed and
plug-flow formats in many different unit process objects. The
biological model is the same, however, regardless of the hydraulic
implementation.
Before discussing
specific models, some common features are presented:
The oxygen transfer
model is based on theory presented in the USEPA Design Manual for
Fine Pore Aeration Systems (USEPA, 1989) and Mueller et al.
(2002). The GPS-X aeration model is suitable for accurate
design of diffused and surface mechanical aeration systems.
In GPS-X, oxygen
transfer to the bulk liquid phase of a biological reactor is
modelled using a dynamic mass balance written for dissolved oxygen
gas. For example, a dissolved oxygen mass balance around a completely
stirred tank reactor (CSTR) is shown below in Equation
6.1.
Equation 6.1
where:
V =
reactor volume (m3)
CL
= concentration of dissolved oxygen (DO) in the reactor (mg/L)
Q = influent
flow rate (m3/d)
Cin =
concentration of DO entering reactor (mg/L)
KLa = oxygen mass
transfer coefficient at field conditions (1/day)
C*∞
= DO saturation concentration at field conditions (mg/L)
r
= rate of use of DO by biomass (g/day), the respiration rate
The volume flows,
and reaction rates are known from specifications or other modelling
equations leaving two terms that must be calculated in order to
solve the dissolved oxygen mass balance over time for the DO
concentration in the reactor, CL:
1.
DO saturation concentration at field conditions,
C*∞, and,
2.
Oxygen mass transfer coefficient at field conditions,
KLa
Calculation of DO Saturation Concentration at Field
Conditions
The DO saturation
concentration at field conditions is calculated as follows:
Equation 6.2
where:
τ
= temperature correction factor (unitless)
β
= correction factor for salts, particulates, and surface-active
substances (unitless)
Ω
= pressure correction factor (unitless)
C*∞20 = DO
saturation concentration at 20°C and 1 atm (mg/L)
The correction factors are used to adjust the DO saturation
concentration to account for the temperature of the liquid, the
pressure at the submergence level of the diffusers, and the salts,
precipitates, and surface-active substances found in the
wastewater.
The temperature
correction factor is calculated as follows:
Equation 6.3
where:
C*st
= surface DO saturation concentration at temperature of t
and 1 atm (mg/L)
C*s20 =
surface DO saturation concentration at 20°C and 1 atm (mg/L)
The surface DO
saturation concentration at liquid temperature t and a
pressure of 1 atm is obtained using a lookup table in GPS-X that is
based on temperature. The lookup table data were taken from
Appendix C of the USEPA Design Manual – Fine Pore
Aeration Systems (USEPA, 1989). When the
temperature falls between two data points in the table, GPS-X uses
linear interpolation to determine the
C*st value. The value of
C*s20 is 9.09 mg/L.
The correction
factor for salts, particulates, and surface-active substances
β, is a parameter that must be
measured or estimated for the wastewater of interest. In
GPS-X, a default value of 0.95 is used.
The pressure correction factor
is calculated as shown below:
Equation 6.4
where:
Pb
= barometric pressure at elevation and air temperature (kPa)
Ps
= standard barometric pressure (101.325 kPa)
pv
= vapour pressure of water at liquid temperature (kPa)
pde
= effective pressure at depth of diffuser submergence
The barometric
pressure at elevation and air temperature is calculated using the
following formula taken from Appendix B-2 of Metcalf
and Eddy (2003):
Equation 6.5
where:
g =
acceleration due to gravity (9.81 m/s2)
M =
molecular weight of air (28.964 kg/kg-mole)
R =
universal gas constant (8314 m/kg-mole K)
Tair = air
temperature (K)
zi
= elevation at position i (m)
The vapour pressure
of water at the liquid temperature is determined using the
Antoine equation (Felder & Rosseau, 1986):
Equation 6.6
where:
A, B, C, = Antoine
coefficients
(found in System > Input Parameters >Physical
form in GPS-X under Physical Constants)
T
= wastewater temperature (°C)
The effective pressure at the depth of the diffuser submergence is
calculated using the following formula:
Equation 6.7
where δ is the depth
correction factor for oxygen saturation and is given by:
Fine Pore and
Jets
Equation 6.8
Coarse
Bubble
Equation 6.9
The parameter d is the depth of submergence of the
diffusers.
The DO saturation concentration at 20°C and 1 atm is calculated
as follows:
Equation 6.10
Oxygen Mass Transfer
Coefficient at Field Conditions Calculation
The value of the
oxygen mass transfer coefficient at field conditions,
KLa, specifies the amount of oxygen supplied to
the aeration tank given the driving force,
and the tank volume. GPS-X provides four ways for the user to
specify the KLa:
1.
If known, the user can directly supply the KLa at
20°C (no alpha,
fouling, or temperature correction) for both diffused and
mechanical aeration. GPS-X converts the KLa to
field conditions.
2.
In the case of mechanical aerators, the user can supply an aeration
power and mechanical aerator oxygen transfer rate from which GPS-X
calculates the KLa.
3.
In the case of diffused aeration, the user can supply an air flow
rate (at either standard or field conditions) and a standard oxygen
transfer efficiency (SOTE) from which GPS-X calculates the
KLa at field conditions.
Alternatively, the user can configure a DO controller which will
manipulate the field KLa directly to match the
desired DO concentration. The availability of these options
depends on the selected aeration method. For diffused
aeration the user can enter airflow, enter KLa,
or use a DO controller. For mechanical aeration the user can
enter power, enter KLa, or use a DO
controller. The four choices for specifying
KLa are discussed in more detail in the following
four sections.
Entering
KLa Directly – Diffused Aeration
The
KLa at field conditions is calculated using
Equation 6.11:
Equation 6.11
where:
KLaT
= mass transfer coefficient at temperature T in °C (1/day)
KLa20
= mass transfer coefficient at 20°C
θ
= temperature correction factor (default value in GPS-X is
1.024)
α
= wastewater correction factor
for KLa20
F
=diffuser fouling factor
(default value in GPS-X is 1.0)
T
= wastewater temperature (°C)
The
αcorrection factor can be specified along the
length of the aeration tank in the case of a plug flow unit
process.
GPS-X calculates other useful
process variables from the standard and field
KLa as detailed below:
Oxygen Transfer
Rate (OTR) at Field Conditions in g/d
Equation 6.12
Standard Oxygen
Transfer Rate (SOTR) in g/d
Equation 6.13
Airflow at
Standard Conditions in m3d
Equation 6.14
where:
SOTE
= standard oxygen transfer efficiency (as a fraction)
CF1
= conversion factor to account for the density, molecular weight,
and O2 mole fraction of the standard air
(U.S. Standard = 277.6533841; European Standard = 300.495893)
In the case of
user-defined standard air (see Entering Airflow
section below), the airflow is calculated using Equation
6.15.
Equation 6.15
where:
molfrO2
= mole fraction of O2 in user-defined air
(mole/mole)
MWO2
= molecular weight of O2 (32 g/mole)
Puser
= density of user-defined air (g/m3)
MWuser
= average molecular weight of user-defined air (g/mole)
The airflow at either U.S. or European standard conditions is
converted to field conditions using the ideal gas law:
Equation 6.16
where:
Tstandard
= air temperature at standard conditions (°C)
Tfield
= air temperature at field conditions (°C)
Equation 6.16 assumes that the field air has the same
humidity as the standard air. If the humidity of the field
air is different than for the standard air, the user can multiply
the GPS-X calculated field air by the ratio of the mole fraction of
oxygen in the field air by the mole fraction of air in the standard
air.
No conversion to
field conditions is undertaken for user-defined standard air.
Entering KLa
Directly – Mechanical Aeration
Mechanical surface
aeration is calculated similarly to diffused; however, the fouling
and depth correction factors are not required and oxygen transfer
is related to the mechanical power input.
For mechanical
surface aeration, the formula for the DO saturation concentration
at field conditions is modified as follows:
Equation 6.17
as δ
= 1 in this case,
the pressure correction factor is re-defined as:
Equation 6.18
The
KLa at field conditions is calculated using
Equation 6.19.
Equation 6.19
The OTR is calculated using Equation 6.20 as shown
below:
Equation 6.20
The SOTR is
calculated as follows:
Equation 6.21
The mechanical power is calculated using the SOTR:
Equation 6.22
where:
Pmechanical
= mechanical power (kW)
η
= mechanical aerator oxygen transfer rate (1.75 kg O2/kW h)
CF2
= conversion factor (24,000)
The default α value for mechanical aeration in GPS-X is
0.9.
The airflow can be
entered at either standard or field conditions. At standard
conditions, the user has three options:
1.
U.S. Standard Conditions:
a.
Temperature = 20°C
b.
Pressure = 1 atm
c.
Relative humidity = 36%
2.
European Standard Conditions:
a.
Temperature = 0°C
b.
Pressure = 1 atm
c.
Relative humidity = 0%
3.
User-Defined Standard Conditions:
The user must specify the properties (mole fraction of O2 in air,
density, molecular weight, and exponent in blower power
equation). When using the user-defined standard conditions
option, GPS-X uses a Henry’s law correction to adjust the value of
C*st.
Equation 6.23
GPS-X converts the airflow to U.S. standard conditions and then
calculates the SOTR using Equation 6.13. The OTR is
calculated using Equation 6.24.
Equation 6.24
The oxygen mass transfer coefficient at field conditions is
calculated using Equation 6.11. If the airflow is
entered at U.S. or European standard conditions, the airflow at
field conditions is calculated using Equation
6.16.
Entering Mechanical
Power
If mechanical power
is entered, the SOTR in g/d is calculated using Equation
6.21. The KLa at field conditions is then
calculated using Equation 6.19 and the OTR is
calculated using Equation 6.20.
DO Control
The user has the
option of controlling the aeration supply to the reactor to
maintain a specific dissolved oxygen setpoint. In this case the
model will calculate the necessary oxygen mass transfer coefficient
at field conditions to maintain the DO at the setpoint.
For diffused
aeration with DO control, GPS-X uses Equation 6.12,
Equation 6.13, and Equation 6.14 to
calculate the SOTR, OTR, and airflow. For mechanical aeration
with DO control, GPS-X uses Equation 6.20,
Equation 6.21, and Equation 6.22 to
calculate the OTR, SOTR, and mechanical power
Standard Oxygen Transfer
Efficiency (SOTE)
The user has the option of using a constant SOTE or using a
correlation in GPS-X to calculate the SOTE. When using the
constant SOTE option, the default value (0.3) represents a typical
estimate of the efficiency for fine bubble diffusers submerged at
4.3 m (14 ft.).
Alternatively, GPS-X has SOTE correlations for fine bubble, coarse
bubble, and jet diffusers. For fine bubble aeration, the SOTE
correlations are based on a regression equation developed by
Hur (1994) which is shown below:
Equation 6.25
where:
AF = air flow per
diffuser (scfm per diffuser)
d =
diffuser submergence (ft.)
DD =
diffuser density (diffusers/100ft2)
A1, A2, A3, A4,
and A5 = regression
parameters
As shown, this equation depends on the depth of submergence of the
diffusers, the diffuser density, and the airflow per
diffuser. Using this equation and the data found in
Hur (1994), Hydromantis re-estimated the regression
parameters to improve the fit of the regression. Separate
regressions were performed for ceramic disc, ceramic dome, membrane
disc, and membrane tube diffusers. The values of the regression
coefficients can be found in the System > Input Parameters
> Physical > More form.
There is also an option to use a user-defined SOTE correlation. The
correlation has the same form as Equation 6.25. The
adjustable regression coefficients are found in the System
> Input Parameters > Operational > More form of
most biological objects.
A separate correlation for coarse bubble and jet diffusers has been
developed by Hydromantis based on data found in Mueller
et al (2002) as shown in Equation 6.26.
Coarse Bubble
and Jet SOTE
Equation 6.26
The SOTE for coarse
bubble and jet diffusers is dependent upon the airflow per diffuser
and the diffuser submergence.
In all the SOTE
correlations, the airflow per diffuser has been limited to be
within the ranges used for calibration. The correlations are
empirical and should not be extrapolated outside of the calibration
range. The ranges are used as follows:
Fine
Bubble
Ceramic
Disc:
0.5 ≤ AF ≤ 5m3/h/diffuser
Ceramic
Dome:
0.5 ≤ AF ≤ 4.4 m3/h/diffuser
Membrane
Disc:
0.8 ≤ AF ≤ 5 m3/h/diffuser
Membrane
Tube: 0.8 ≤
AF ≤ 15 m3/h/diffuser
Coarse Bubble
14 ≤ AF ≤ 58
m3/h/diffuser
Jet
8 ≤ AF ≤
140 m3/h/diffuser
GPS-X applies a warning message to the log window if the airflow
per diffuser is outside the ranges given above.
SOTE in Deep Tanks
The fine bubble
SOTE correlation is modified for tanks deeper than 8 m. Work
by Pöpel and Wagner (1994) has shown that the SOTE does not
continue to increase linearly with the depth of submergence beyond
depths of approximately 8 m (see Figure 8 in Pöpel and Wagner
(1994)). At depths greater than 8 m the SOTE
starts to level off so that the SOTE can never be greater than 100
%.
An extra quadratic
term in diffuser submergence is added to Equation
6.25 when the diffuser submergence is greater than 8 m, as
shown in Equation 6.27.
Equation 6.27
The regression
parameter A6 can be accessed in GPS-X in
the
Input Parameters > Operational > More
form. The parameter A6 was estimated by
Hydromantis uses linear regression to fit Equation
6.26 to the solid curve in Figure 8 of Pöpel and Wagner
(1994). In order to perform the regression it was
necessary to first adjust the airflow per diffuser and the diffuser
density to match the SOTE of 20 % at depth of 5 m shown in Figure 8
of Pöpel and Wagner (1994).
The deep tank SOTE
correlation is only applied for fine bubble diffusers. It can
be turned off by setting the parameter A6 to zero
in GPS-X.
Blower Wire Power
The wire power in
kW consumed by the blowers in order to deliver the required air is
calculated in GPS-X as follows:
Equation 6.28
where:
DP = delivered power of
blowers (kW)
e =
overall efficiency of mechanical equipment (i.e. blowers, motors,
coupling, and gearbox)
The overall
efficiency of the mechanical equipment is entered into GPS-X in the
Input Parameters > Operating Cost form of
most biological objects. The delivered power of the blowers is
calculated using the adiabatic compression equation (Mueller
et al., (2002)):
Equation 6.29
where:
w =
mass flow rate of air
R =
universal gas constant
Ta
= blower inlet air temperature (degrees K)
K =
R/Cp where Cp is the heat
capacity of air at constant pressure
(K = 0.283 for U.S Standard Air and K = 0.2857 for
European Standard Air)
Pd
= absolute pressure downstream of blower (discharge pressure in
kPa)
Pa
= absolute pressure upstream of blower (inlet pressure in kPa)
The discharge pressure of the blower is calculated as follows:
Equation 6.30
where:
Ps
= barometric pressure (default is 101.325 kPa)
g =
acceleration due to gravity (9.81 m/s2)
Δpd = pressure drop
in piping and diffuser downstream of blower (found in Input
Parameters > Operating Cost form of most biological
objects)
The inlet pressure is calculated as shown in Equation
6.31
Equation 6.31
where:
Δpa = pressure drop
in inlet filters and piping to blower (found in
Input Parameters > Operating Cost form of
most biological objects.)
This section
discusses the aeration forms that are applicable to most biological
objects. The forms from the Completely-Mixed Tank are
used as an example.
Aeration Setup Form
The Aeration
Setup form is accessed by right-clicking on most biological
objects and selecting Input Parameters >
Operational. Figure 6‑1 shows the GPS-X form.
The parameters are discussed below:
Aeration method
(Diffused Air/Mechanical (Surface Aeration)): Specifies
whether the aeration system is diffused or mechanical
(surface).
Specify oxygen
transfer by… (1 Entering Airflow/2 Entering KLa/3
Entering Mechanical Power/4 Using a DO Controller):
Specifies the method of oxygen transfer, entering airflow,
KLa, mechanical power, or using a DO
controller. Entries that do not apply are greyed out. For
example, you cannot specify the oxygen transfer by entering
mechanical power if the aeration method is diffused.
Oxygen
mass transfer coefficient (clean water): Clean water
KLa at 20°C with no alpha or fouling
factor correction.
Air flow into
aeration tank: Diffused air flow into the aeration tank.
The airflow can be at standard conditions (U.S., European, or
User-Defined) or at field conditions depending on user selections.
The user can select standard or field conditions in the Input
Parameters > Operational > More… form. The user
can select which standard conditions are used in the
Input Parameters > Physical >More…
if local conditions for O2 solubility are used, or in
the System >Input Parameters > Physical form if
global conditions for O2 solubility are used.
Aeration
power: The power of the mechanical aeration system.
Figure 6‑1 - Aeration Setup Form in Operational
Form
By clicking on the
More… button, the form shown in Figure 6‑2 and
Figure 6‑3can be accessed. The parameters found
in this form are discussed in the subsequent sections.
Figure 6‑2 - General Aeration Setup >
More... Form
Figure 6‑3 - Diffused Aeration Setup >
More... (Part 1)
Figure 6‑4 -
Diffused Aeration Setup > More... Form (Part 2)
General Sub-Section (Under
More…)
Beta factor (for
DO saturation): Specifies the DO saturation concentration
correction factor for sales, particulates, and surface-active
substances. The default value is 0.95.
Temperature
coefficient for KLa: Specifies the
temperature correction factor for KLa, θ
(default value is 1.024).
Sub-Section (Under
More…)
Minimum oxygen
mass transfer coefficient: Specifies the minimum value of
the KLa at field conditions. GPS-X bounds
the field KLa to be greater than or equal to this
maximum in all cases.
Maximum oxygen
mass transfer coefficient: Specifies the maximum value of
the KLa at field conditions. GPS-X bounds
the field KLa to be less than or equal to this
maximum in all cases.
Mechanical (Surface
Aeration) Sub-Section (Under More…)
Mechanical
aerator oxygen transfer rate: Specifies the mass of
oxygen transferred to the liquid per unit of energy used by the
mechanical aerator.
Alpha factor
(mechanical aeration): Wastewater correction factor for
mechanical KLa. The default value is
0.9.
Diffused Air Sub-Section
(Under More…)
Input air flow
at… (Field Conditions/Standard Conditions): Allows user
to specify whether the diffused air flow is entered at standard
conditions or field conditions. The user can select which
standard conditions are used in the
Input Parameters > Physical > More … if local
conditions for O2 solubility are used or in the
System > Input Parameters > Physical form if
global conditions for O2 solubility are used.
Diffuser type
(Fine Bubble/Coarse Bubble/Jet/User-Defined): Specifies
the type of diffuser to be used. This option is only active
if diffused air has been selected. GPS-X uses this selection
to select the appropriate alpha value and to select the appropriate
SOTE correlation (if applicable).
Alpha factor
(fine bubble): Specifies the value of the wastewater
correction factor for the fine bubble KLa.
The default value is 0.6.
Alpha factor
(coarse bubble): Specifies the value of the wastewater
correction factor for the coarse bubble
KLa. The default value is 0.8.
Alpha
factor (jet): Specifies the value of the wastewater
correction factor for the jet KLa. The
default value is 0.85.
Alpha factor
(user-defined): Specifies the value of the wastewater
correction factor for the user-defined KLa.
The default value is 0.6
Fouling
constant: Diffuser fouling factor. Default value is 1
(i.e. no fouling)
Depth correction
factor for user-defined (Fine Bubble/Coarse Bubble): When
the user-defined diffuser type is being used, this parameter
selects whether Equation 6.8 or Equation
6.9 is used to calculate the depth correction factor for
the DO saturation concentration.
Oxygen Transfer Efficiency
(SOTE) Sub-Section (Under More…)
SOTE type
(Constant/Correlation): Specifies whether the SOTE is set
by the user or calculated by a correlation. The correlation
used depends on the diffuser type selected.
Standard oxygen
transfer efficiency: When the SOTE type is set to
constant, this value is used for the SOTE.
Diffuser
submergence: Specifies the depth of submergence of the
diffusers.
Method of
specifying diffuser setup (Enter Number of Diffusers/Enter Diffuser
Density): Allows the user to either specify the number of
diffusers or the diffuser density (diffuser area/tank area).
It is important to enter the diffuser density or the number of
diffusers if an SOTE correlation is being used or the user requires
GPS-X to display accurate values of the airflow per diffuser and
the number of diffusers per unit area.
Diffuser density
(diffuser area/tank area): When the method of specifying
the diffuser setup has been selected as diffuser density, this
parameter allows the user to enter the diffuser density as a
fraction or a percentage. Diffuser densities between 5 % and
25 % are typical.
Number of
diffusers or jets: When the method of specifying the diffuser
setup has been selected as enter number of diffusers, this
parameter allows the user to enter the number of diffusers.
Tank floor
area: The area of the tank floor. It is important to
enter the tank foor area if an SOTE correlation is being used or
the user requires GPS-X to display an accurate value of the number
of diffusers per unit area.
Area per
diffuser: The floor area taken up by a diffuser. It is
important to enter the area per diffuser if an SOTE correlation is
being used and the user is entering the diffuser density. The
default value is 0.038 m2 which is typical for a 9″
diameter ceramic diffuser.
Fine bubble
diffuser head type (Ceramic Disc/Ceramic Dome/Membrane Disc/
Membrane Tube): Used to specify the fine bubble diffuser head
type when fine bubble diffused aeration has been specified and an
SOTE correlation is being used.
User-Defined SOTE
Regression Coefficients Sub-Section (Under More…)
SOTE regression
constant A1 (user-defined): Regression parameter
A1 in user-defined SOTE correlation.
SOTE regression
constant A2 (user-defined): Regression parameter
A2 in user-defined SOTE correlation.
SOTE regression
constant A3 (user-defined): Regression parameter
A3 in user-defined SOTE correlation.
SOTE regression
constant A4 (user-defined): Regression parameter
A4 in user-defined SOTE correlation.
SOTE regression
constant A5 (user-defined): Regression parameter
A5 in user-defined SOTE correlation.
Deep Tank User-Defined
SOTE Regression Coefficient (Under More…)
Deep Tank SOTE
regression constant A6 (user-defined): Regression
parameter for the deep tank SOTE term in the user-defined SOTE
correlation.
Physical Form – Local
Conditions for O2 Solubility (Under More…)
When local
conditions are being used for oxygen solubility, the physical data
related to the aeration model is accessed by right-clicking on most
biological objects and selecting Input Parameters >
Physical > More …. The available parameters
are shown in Figure 6‑5 and are detailed below.
Figure 6‑5 - Physical > More... Form within
an Object
Local Environment
Selection Sub-Section
Use local
settings for O2 solubility and biological activity (ON –
OFF): Selects whether local or global conditions are used
for oxygen solubility.
Oxygen Solubility (if
individual settings are used) Sub-Section
Liquid
temperature: Allows the user to specify the local
wastewater temperature
Blower inlet air
temperature: Allows user to specify the inlet air
temperature for the blowers.
Elevation above sea level: Specifies the elevation
above sea level.
Standard
air conditions (U.S. (air temp 20°C, 36% humidity)/ European
(air temp 0°C, 0%
humidity/User-Defined): If the aeration method is
diffused, this switch selects which standard conditions are used to
calculate the standard air flow used within GPS-X and displayed as
an output. If user-defined standard air is selected, the user
must enter the properties of their air in the Properties of
User-Defined Air sub-section. The user-defined air option
provides the user with a method of specifying pure oxygen.
Properties of User-Defined
Air Sub-Section
Mole fraction of oxygen in user-defined air:
When user-defined standard air has been selected, this parameter
specifies the oxygen mole fraction of the user-defined air.
The default value is 1 which is consistent with pure oxygen.
Density of user-defined air: When user-defined
standard air has been selected, this parameter specifies the
density of the user-defined air. The default value is 1,429
mg/L which is consistent with pure oxygen.
Molecular weight of user-defined air: When
user-defined standard air has been selected, this parameter
specifies the molecular weight of the user-defined air. The
default value is 32 g/mole, which is consistent with pure
oxygen.
Exponent in
blower power equation: When user-defined standard air has
been selected, this parameter specifies the value of exponent
K in Equation 6.28 for the user-defined air. The
default value is 0.284, which is consistent with pure oxygen.
Physical Form – Layout
Wide Settings
When global
conditions are being used for oxygen solubility or it is necessary
to change global physical constants important to aeration, the
relevant physical data is accessed by selecting System >
Input Parameters > Physical Environment
Settings. The available parameters are shown
in Figure 6‑6 and are detailed below.
Oxygen Solubility
(Layout-Wide Settings) Sub-Section
Most of the parameters in this sub-section of the layout-wide
physical form are duplicates of the local versions discussed
earlier. The one exception is as follows:
Barometric
pressure at sea level: Specifies the pressure at sea
level. The default value is 101.325 kPa.
Properties of User-Defined
Air Sub-Section
The parameters in
this sub-section of the layout-wide physical form are duplicates of
the local versions discussed earlier.
SOTE Regression
Coefficients Sub-Section
By clicking on the
More… button in the Properties of User-Defined Air section,
the user can specify the regression coefficients used in the
available SOTE correlations.
Figure 6‑6 - Physical Environment Settings
(Layout-Wide Settings)
Molecular weight of air (@ U.S. Standard Conditions):
Specifies the molecular weight of air at U.S. Standard Conditions
which is used in Equation 6.5 to calculate the
barometric pressure at the elevation and air temperature
specified. GPS-X does its internal calculations in U.S.
Standard Conditions.
Gas constant: Specifies the value of the universal gas
constant.
Antoine coefficient A1: Coefficient A in the
Antoine equation which is used to calculate the vapour pressure of
water at a given water temperature.
Antoine coefficient A2: Coefficient B in the
Antoine equation which is used to calculate the vapour pressure of
water at a given water temperature.
Antoine coefficient A3: Coefficient C in the
Antoine equation which is used to calculate the vapour pressure of
water at a given water temperature.
Blower Cost Sub-Section
(Operating Cost Form)
The blower cost
sub-section can be accessed by right-clicking on most biological
objects and selecting Input Parameters > Operating
Cost. The GPS-X form is shown in Figure
6‑7. The available parameters are described
below.
Combined blower/motor efficiency: Specifies the overall
efficiency of the mechanical aeration equipment (i.e. blowers,
motors, coupling, and gear box). The default value is 0.7 or
70 % efficiency.
Pressure drop in inlet filters and piping to blower:
Specifies the pressure drop in the inlet filters and piping to the
blowers for the purposes of calculating the wire power supplied to
the mechanical aeration equipment. The default value is 1
kPa.
Pressure drop
in piping and diffuser downstream of blower: Specifies
the pressure drop in the piping and diffusers downstream of the
blowers for the purposes of calculating the wire power supplied to
the mechanical aeration equipment. The default value is 7
kPa.
Figure 6‑7 - Blower Cost Form
The aeration output
variables can be accessed by right-clicking on most biological
objects and selecting Output Variables > Oxygen
Transfer. For single tank objects, the same form can
be accessed by right-clicking on the overflow stream. For
muli-tank objects such as the plugflow tank, the total airflow,
SOTR, OTR, and mechanical power can be accessed by right-clicking
on the overflow stream and selecting Total Air Flow,
Total Oxygen Transfer, or Total Mechanical Power.
The output
variables associated with the blower calculations can be accessed
by right-clicking on the object (single tank objects only) or the
overflow stream and selecting
Output Variables > Operating Cost.
MBR Aeration Model
In the Membrane Bioreactor and Completely-Mixed MBR
objects the biological air and cross-flow air in the membrane tanks
are tracked separately. The biological air is handled
similarly as in the other objects in GPS-X with the exception that
surface mechanical aeration is not an available option.
The cross-flow air is assumed to be delivered using a coarse bubble
aeration system. The user can enter the following variables
for the cross-flow air: air flow, alpha, and SOTE. The
cross-flow air is assumed to be at the conditions specified for the
biological air (i.e. Standard or Field). Unlike
other objects in GPS-X, the user can specify whether the air is at
standard or field conditions even if the user is not entering the
biological airflow. GPS-X tracks the biological and
cross-flow air separately but also calculates the total air flow,
SOTR, and OTR delivered to the tank.
In the MBR objects,
lower default alpha values are used to reflect the high MLSS
concentrations typically found in MBRs.
This model extends
the capability of the standard aeration model in activated sludge
unit process models to simulate the head-loss in the air delivery
model, thus simulating the dynamic changes in the blower discharge
pressure. The air delivery head loss model can be used to
simulate the following losses in air delivery system.
1)
Air diffuser
2)
Air delivery pipes and
fittings
3)
Control valve
In the air delivery
head loss model, the required air flow rate in each biological
reactors (user specified or calculated by DO controller) are used
to estimate head-losses in various components of the air delivery
system. To model the air delivery system, the model uses a
predefined air delivery system as shown below. The model is based
on the proposed model by Gray and Kestel, 2013 with some
modifications.
Figure 6‑8 -
Predefined air delivery system
The pressure loss
calculations for the diffuser, air pipe and control valve are
described in following sections.
Diffuser headloss
The diffuser head
loss is calculated based on a diffuser head loss curve. The head
loss curve is defined by the user by providing a number of points
specifying air flow rate per diffuser and the expected head loss.
The air flow rate per diffuser is calculated based on the air flow
rate requirement and the number of diffusers in each tank set by
the user in the aeration menu. The user may use available diffuser
head loss curves from the aeration system supplier. The default
diffuser head loss curve in model is as shown below.
Figure 6‑9 - Default curve for diffuser
head loss
User may refer to
ASCE (1988) reference on Aeration for the expected diffuser head
loss from different diffuser systems.
Air delivery Pipe Head Loss
The head loss in a
straight pipe is calculated using the following equation (Qasim,
1999).
Equation 6.32
Where:
- head
loss, m H2O
f
- friction
factor,-
L
- length of
pipe,m
D
- pipe diameter,
m
Qair - air flow
are, m3/min
P
- air supply
pressure, atm
T
-
temperature in pipe, oK
The friction factor
is calculated using the following equation:
Equation 6.33
The temperature in
pipe is estimated as below:
Equation 6.34
Where:
To
- ambient air temperature, oK
Po
- ambient barometric pressure, atm
Minor Fitting Losses
The minor fitting
losses in the ell, tees and other fittings are estimated using the
pipe head loss equation as above. The user is required to estimate
the equivalent length of the fittings and input it into the model
(Qasim, 1999).
The user may use
the following equation to find the equivalent length of the
fittings.
Equation 6.35
Where:
L -
Equivalent length of pipe fitting, m
C
- C value for
equivalent pipe length
D -
Diameter of fitting, m
The C values for
the fitting are typically available from the
manufacturer/supplier.
Valve Head loss
The head loss in a
valve is estimated based on the following equation used by Gray and
Kestel, (2013).
Equation 6.36
Where:
Q
- air flow rate,
scfm
Pi
-inlet pressure, psia
Po
-outlet pressure, psia
T -
temperature, oK
Sg
- specific gravity of air,-
CV -
valve flow coefficient,
The specific
gravity of air is estimated at the pressure and temperature of air
in pipe as estimated for the pipe head loss. The valve flow
coefficient depends on the valve position and the valve
characteristics curve. The model uses linear, equal
percentage, hyperbolic, exponent and square root valve
characteristic curve equations to estimate the valve flow
coefficient. The equations used are as below.
Linear
Equation 6.37
Equal
Percentage
Equation 6.38
Hyperbolic
Equation 6.39
Exponent
Equation 6.40
Square
Root
Equation 6.41
Where:
CV - valve
flow coefficient, scfm/psia
CV,max - valve flow coefficient at
fully open valve, scfm/psia
VP - valve
position (0 – 1), -
VR - valve
factor for the curve, -
m -
exponent, -
For a given air
flow rate in each reactor, the model estimates the head loss in
diffusers, drop pipes and fittings and control valves. Based on the
hydrostatic pressure and the calculated head losses, the pressure
at point P1 and P2 (Figure 6‑8) are estimated. The pressure
at point P1 is balanced by manipulating the valve positions. Two
algorithms of Most Open Valve (MOV) and Set Point Pressure are
implemented. In MOV algorithm, the valve corresponding to the
maximum head loss is opened until it is open to specified maximum
and then the other valves are adjusted to achieve the same pressure
as estimated by the MOV. In the Set Point Pressure method, all the
valves are adjusted to achieve the set point blower pressure.
State Variables
A few comments
about the state variables used in each of these models is required.
For all the models, the symbolic name of the state variables is
prefixed with either an X for particulate component, S for soluble
component, or in some cases a G for gaseous component. This
designation follows the IWA (Henze et al.,
1987a) convention, but is somewhat arbitrary because some
soluble and particulate components are colloidal and will pass
through a 0.45mm
filter but still be classified as part of a particulate
component.For example the slowly biodegradable material, designated
Xs, may include soluble and/or colloidal material. This approach greatly
facilitates the model development; however it invariably introduces
some error into the models. The issues of wastewater
characterization are important and the modeller should refer to the
papers listed with the individual models discussed below.
Model Matrix
Most biological process models now follow a standard matrix
format. An example of this format is shown in Table
6‑1.
Table 6‑1 –
Example Model Matrix (Wentzel et al., 1987a)
In this table, the
components or state variables of the Monod-Herbert
model (Herbert, 1958), designated by a variable
with a subscript i, are numbered and listed across the top. Three
state variables are defined (Xb, SS and
So), each having its own column. Names and
units for each state variable are provided in the bottom row of
each of these columns. The important processes, designated
pj, in the system, which result in changes in the state
variables, are shown in separate rows; the actual process rate
(kinetic expression or rate equation) is shown in the rightmost
column of each of these rows. All the necessary kinetic parameters
are defined in the lower right-hand corner of the table.
The entries within the table are the stoichiometric parameters or
relations, designated vij, used in defining the net process
rate for a component. These parameters define the mass
relationships between components and are defined in the lower-left
hand corner of the table. If a process does not directly affect a
component's rate, then the table cell will be empty (the entry is
assumed to be zero in this case).
The net reaction
rate of a component, designated ri, is the
sum of all the process rates, which cause a change in the mass of
that component. The expression used to determine the net rate is
listed in the table in the row labeled "Observed Conversion
Rates". When the model is presented in matrix format,
this equation has a simple visual interpretation. To determine the
net rate of change for a component, first identify the column of
the component of interest and move down that column until you find
a table cell containing an entry. Multiply the table cell entry by
the process rate shown in the rightmost column. The sum of these
individual process rates is the net reaction rate. Do likewise for
all remaining rows in the column, which contain stoichiometric
parameters.
Stoichiometric
Parameters
Two common
stoichiometric parameters appearing in the biological models in
this chapter are derived from their respective chemical equations.
First is the nitrogen to oxygen stoichiometric parameter used in
the denitrification equation. A value of 2.86 is used throughout
the GPS-X libraries for all the biological models. This value is
derived from the molecular ratios shown in the equation below:
Equation 6.42
The ratio of mass
oxygen produced per mass nitrogen gas produced is 160:56 or
2.86:1. Similarly, the stoichiometric ratio of oxygen to
nitrogen required for nitrification (used in many of the biological
models) has a value of 4.57 which is derived from the molecular
ratios shown in the equation below:
Equation 6.43
The ratio of mass
of oxygen to mass of nitrogen is 64:14 or 4.57:1.
Since these stoichiometric ratios
cannot change, the values are hard-coded in GPS-X without user
access.
Introduction
The International Association on Water Pollution Research and
Control (IAWPRC) Task Group realized that due to the long solids
retention times and low growth rates of the bacteria, the actual
effluent substrate concentrations between different activated
sludge treatment plants did not vary greatly. What were
significantly different were the levels of MLSS and electron
acceptor (oxygen or nitrate). Thus the focus of the Activated
Sludge Model No. 1 (called asm1 in GPS-X) is the
prediction of the amount and change of the solids and electron
acceptor.
The Task Group considered the trade-off between model accuracy and
practicality. They identified the major biological processes
occurring in the system and characterized these processes with the
simplest rate expressions that could be used, resembling the real
reactions.
The use of switching functions was made by the
Task Group since some reactions depended on the type of electron
acceptor present. These functions were of the form:
Equation 6.44
At low concentrations
of dissolved oxygen (SO), the parameter
KOH dominates the expression and approaches a
value of zero. At high values of SO, the
parameter KOH would be negligible and the
expression approaches unity. If the switching function was
inverted, then the limits when SO were high or
low are reversed. A consequence of using switching functions of
this form is that they are continuous functions unlike
discontinuous on/off switches which are more difficult to
simulate.
Conceptual Model
In the development
of activated sludge modelling, the manner in which the quantity of
organic matter is measured (BOD, COD or TOC) is inconsistent. The
Task Group decided to use COD since mass balances can be carried
out and since it has links to the electron equivalents in the
organic substrate, biomass and electron acceptor.
The organic material is
categorized according to a number of characteristics. First, is the
biodegradability of the
material. The non-biodegradable organics pass through the system
unchanged and can be further categorized according to their
physical state (soluble or particulate), which is removed from the
system by different pathways. The particulate material is generally
removed with the waste activated sludge, while the soluble material
leaves with the effluent. The biodegradable material is categorized
as either readily or slowly biodegradable. The Task Group treated
the former as soluble material, while the latter was treated as
particulate material (this is not strictly correct, but simplifies
matters). The readily biodegradable organics may be utilized for
cell maintenance or growth with a transfer of electrons to the
acceptors. The particulate (slowly) biodegradable substrate is
hydrolyzed to readily
biodegradable material, assuming no energy utilization and no
corresponding use of electron acceptor.
The hydrolysis rate is usually slower than the utilization of
readily biodegradable substrate so that it is the rate limiting
step, if only slowly biodegradable substrate is available.
Two types of biomass are modelled:
1.
heterotrophic; and
2.
autotrophic
The heterotrophic biomass is generated by the growth on readily
biodegradable substrate under aerobic or anoxic conditions and
decays (including endogenous respiration, death, predation and
lysis) under all conditions. The autotrophic biomass is generated
under aerobic conditions only utilizing ammonia for energy and
decays under all conditions.
The nitrogenous material
is categorized according to its biodegradability and physical
state. The non-biodegradable particulate material is modelled as a
fraction of the non-biodegradable particulate COD, while the
non-biodegradable soluble material is ignored. The biodegradable
nitrogenous material is divided into ammonia (free and ionized),
soluble organic and particulate organic. Particulate organic is
hydrolyzed to soluble organic, while soluble organic is converted
to ammonia by the heterotrophic biomass. The conversion of ammonia
to nitrate by the autotrophs is assumed to take place in one
step.
The Model Matrix for
asm1 is found in Appendix A.
Asm2 is no
longer implemented in GPS-X, in favour of using the asm2d
model (see below), which corrects errors and deficiencies from the
original published model.
Introduction
This model (asm2d) is an implementation of the Activated
Sludge Model No. 2d (Henze et al., 1998).
The model structure, default values and all other model aspects
follow the publication in every detail.
This model is an
extension of asm1, primarily to handle biological phosphorus
removal systems. The model matrix is shown with the nomenclature
used in the GPS-X implementation. Users of this model should
consult the reference (Henze et al., 1998) for
details of this model.
The asm2d
model is implemented in the CNP and CNPIP library, and the Model
Matrix is found in Appendix A.
ASM2d Model
Components
One major
difference in the way this model is presented compared with the
other models is seen in its matrix description. The stoichiometric
coefficients for ammonia and soluble phosphorus are listed outside
the table. The reason behind this is that to eliminate the organic
nitrogen state variables (xnd and snd); they are now
incorporated into the other soluble and particulate organic
components as a fixed fraction. Similarly, a fixed fraction of
phosphorus is included in the organic components.
State Variables
Each of these state
variables represents a spectrum of organic biodegradable material.
Another state variable is used to model the other soluble organic
material, which is not biodegradable (si). This material is
part of the influent and can be produced during some hydrolysis
processes. The soluble nitrogen components consist of 1) ammonia
and ammonium (snh); and 2) nitrate and nitrite (sno).
The dinitrogen gas produced (snn) is also modelled, but is
considered insoluble and immediately comes out of solution. Oxygen
(so) and inorganic soluble phosphorus (sp) are the
other two soluble states. The inorganic soluble phosphorus is
typically ortho-phosphate.
Looking at the
particulate states, the model includes three types of biomass:
1.
heterotrophic organisms (xbh);
2.
nitrifying organisms (xba); and
3.
phosphate accumulating organisms (xbp)
The phosphate
accumulating organisms’ state variables do not include the internal
storage products, which are separate state variables. Particulate
non-biodegradable organic material is also modelled (xi).
Although it is not removed from the system, it may be generated
during cell decay. The two internal storage products of the
phosphorus accumulating organisms are:
1.
internally stored COD (xbt); and
2.
poly-phosphate (xpp).
Neither of these
two components is included in the mass of the phosphorus
accumulating organism. The last state variable is the slowly
biodegradable substrates (xs) which must undergo hydrolysis
before available as a substrate. Particulate inorganic inert solids
(xii) are present in the system, but do not interact with
the biological model.
Processes
The processes
described by this model are separated into four groups:
1.
processes involving hydrolysis;
2.
processes involving heterotrophs;
3.
process involving autotrophs; and
4.
processes involving phosphorus accumulators.
The hydrolysis
processes include the aerobic, anoxic and anaerobic hydrolysis of
slowly biodegradable organic material into soluble substrate
(processes 1-3). The rate equations are similar for the three
processes; however, the rate constants under these different
environmental conditions are not well known. The hydrolysis of
organic nitrogen is not explicitly included in this model. Rather,
a fraction of the organic particulate material is assumed to be
organic nitrogen and therefore hydrolyzes at the same rate as the
organic particulate substrate.
The heterotrophic
processes include the aerobic growth on two substrates
(processes 4 and 5), their corresponding anoxic growth
(processes 6 and 7). The fermentation of organic material under
anaerobic conditions is also accounted for by process 8. This
process has been identified as one requiring more research into its
understanding. Since little is known about this process, a large
range of kinetic parameters for this rate may be found during the
modelling exercise. Process 9 models the death and lysis of the
heterotrophs.
Processes 10-15
describe the phosphorus accumulating bacteria:
1.
the internal storage of fermentable products (process 10);
2.
the internal storage of poly-phosphate (process 11);
3.
the aerobic growth of polyP bacteria (process 12); and
4.
the lysis of the bacteria and their corresponding internally stored
products (processes 13-15)
The final two processes in the model are the aerobic growth of the
nitrifiers (process 16) and their death and lysis (process 17). The
kinetic parameters of this model, like those of the mantis
model, are temperature dependent. A similar Arrhenius type function
is used to describe this dependency.
This model is a minor extension of the original asm2 model.
It includes two additional processes to account for phosphorus
accumulating organisms (PAOs) using cell internal storage products
for denitrification. Whereas the asm2 model assumed PAOs to
grow only under aerobic conditions, the asm2d model includes
denitrifying PAOs.
The Activated
Sludge Model No. 3 (Gujer et al., 1999)
relates to ASM1 and corrects for some inadequacies of ASM1. The
main features of the model are:
·
Hydrolysis is independent of the electron donor, and occurs at the
same rate under aerobic and anoxic conditions.
·
Lower anoxic yield coefficients are introduced.
·
Decay of biomass is modelled as endogenous respiration (vs. the
“death regeneration” concept used in asm1).
·
Storage of COD by heterotrophs under anoxic and aerobic conditions
is modelled.
·
It is possible to differentiate between anoxic and aerobic
nitrifier decay rates.
·
Ammonification of SND and hydrolysis of biodegradable, particulate
nitrogen (XND) are omitted. Instead, a constant composition
of all organic components has been assumed (constant N to COD
ratio).
·
Alkalinity limitation on the process rates is considered.
Figure
6‑10 shows a schematic of the processes simulated in the
asm3 model.
Figure 6‑10 - ASM3 Model Processes
The Model Matrix
for asm3 is found in Appendix A.
The mantis
model is identical to the IAWPRC Activated Sludge Model No. 1
(asm1), except for the following modifications:
·
Two additional growth processes are introduced, to allow for growth
of heterotrophic biomass with nitrate as a nutrient.
·
Switching functions for nitrogen as a nutrient (and phosphorus, in
applicable libraries) and alkalinity for growth
·
Separate half-saturation coefficients for oxygen for aerobic and
anoxic growth, to allow for calibration of simultaneous
nitrification/denitrification.
The additional growth processes account for the observed growth of
organisms during conditions of low ammonia and high nitrate. Under
these conditions, the organisms can uptake nitrate as a nutrient
source.
The temperature
dependence of the kinetic parameters is described by an Arrhenius
equation. See Appendix A for the Model matrix describing this
model.
Aerobic denitrification is
included in the model according to the Münch modification
(Münch et al., 1996). In many cases
modellers have seen nitrate levels overpredicted in their models
due to the simplifications in spatial resolution (ideally mixed
aeration tanks, no oxygen diffusion limitation in floc cores,
etc.). The new modification, consisting of one new anoxic oxygen
half-saturation coefficient makes anoxic growth rates adjustable
independently from aerobic growth, and the coefficient itself is an
indication of the degree of aerobic denitrification occurring
within the plant modelled. The default value of the new constant is
set equal to the aerobic oxygen half saturation.
Introduction
Dold's
general model is not implemented in GPS-X, in favour of the
newgeneral model. However, a description of the
general model (which makes up the basis for the
newgeneral model) is presented here.
In the following
sections, the general model of Dold (1990) is
described. This model was derived from a combination of the asm1
model for non-polyP heterotrophic organisms and autotrophic
organisms (Henze et al., 1987a, 1987b) and the
Wentzel et al. (1989b) model for polyP
organisms.
General Model Components
(Non-PolyP Organisms)
The general
model component proposed for describing the kinetic response of the
non‑polyP heterotrophic and autotrophic organism masses is based on
the asm1 model (Henze et al., 1987a,
1987b), with three modifications/extensions:
1.
The nitrogen source for cell synthesis
2.
Conversion of soluble readily biodegradable COD to short-chain
fatty acids (SCFA’s).
3.
Growth of non-polyP heterotrophs on SCFA.
Nitrogen Source for Cell
Synthesis
In reviewing the
initial asm1 model version, Dold and Marais (1986)
postulated that under certain circumstances, nitrate, instead of
ammonia nitrogen, may serve as the nitrogen source for cell
synthesis purposes. This postulate was confirmed from analysis of
data collected over an extensive period, particularly in multiple
series reactor configurations operated at long sludge ages and
which exhibited high nitrification rates. The use of nitrate as a
nitrogen source for polyP organism synthesis, when the ammonia
concentration dropped to low levels, was also observed by
Wentzel et al., (1989b). On the basis of this
information, an additional two processes have been incorporated
into the asm1 model version to give four growth processes:
aerobic and anoxic growth of non-polyP heterotrophs with either
ammonia or nitrate as the N source for synthesis.
Growth of non-polyP
heterotrophs on SCFA
In the asm1
model, readily biodegradable soluble COD is utilized by the
non-polyP organisms (i.e. heterotrophs) in four possible growth
modes - under aerobic or anoxic conditions with either ammonia or
nitrate as the nitrogen source for synthesis purposes. For
biologically-enhanced phosphorus removal (BEPR) systems it is
necessary to distinguish between "complex" and SCFA readily
biodegradable COD; therefore, it is necessary to duplicate the four
growth processes in the asm1 model to account for possible
growth on the two components of the readily biodegradable COD for
the mixed culture system. With regard to growth on SCFA it is
likely that only one of the four processes would be of consequence
- anoxic growth with ammonia as the N source. This is because SCFAs
are removed in the unaerated zones at the "front end" of the
continuous flow systems and do not enter the aerobic zones in
appreciable concentrations. However, for completeness all four
growth processes in the asm1 model (for "complex" COD) were
duplicated in the general model for growth of non-polyP
organisms with SCFA as substrate. The same kinetic formulations and
stoichiometry for growth on "complex" readily biodegradable COD and
SCFA have been used.
General Model Components
(PolyP Organisms)
The general
model component proposed for describing the kinetic response of the
polyP heterotrophic organism mass is based on the enhanced culture
model of Wentzel et al. (1989b), with one
additional process for anoxic growth of polyP organisms described
by Dold (1990) (and the one additional stoichiometric
constant, fup, for anoxic growth). It should be noted that
the phenomenon of accumulation of un-biodegradable soluble COD from
endogenous processes in the enhanced cultures also received
attention. Wentzel et al. (1989a) suggested
that this material would be used as a substrate source by non-polyP
organisms in mixed culture systems.
While this is likely, no change was made to the model in this
respect because the amount of generation in the mixed culture
systems is small.
The values of the
kinetic parameters, as with the asm1 model, are those for
20°C.
The general model is no
longer available in GPS-X, in favour of using the newgeneral
model (see below), which corrects deficiencies of the
general model.
New General Extension
The
newgeneral model is based on the general model with
significant changes to account for so-called 'COD losses' that have
been reported in nutrient removal systems.
Some key features
of the newgeneral model include:
·
‘COD’ loss yields
·
Hydrolysis efficiency factors
·
Heterotrophic yield coefficients for different electron acceptor
conditions.
The modelling of
BNR activated sludge systems has identified mass balance problems
in some facilities. That is, several experimental and full-scale
evaluations of BNR facilities have revealed imbalances between what
is going into the plant and what is going out (i.e. less COD is
going out than going in, hence the COD 'loss'). Both from a
theoretical and modelling point of view this causes some problems,
and there is no world-wide accepted explanation for these
imbalances. Nevertheless, these discrepancies must be modelled. To
account for these mass balance problems (referred to as anaerobic
stabilization or COD losses), four COD loss terms are included in
the newgeneral model. These include:
·
Hydrolysis efficiency factor (anoxic)
·
Hydrolysis efficiency factor (anaerobic)
·
Fermentation volatile fatty acid (VFA) yield.
·
PolyP PHB yield on sequestration of VFAs.
The
newgeneral model has proven itself to be a reliable
predictive tool for the BNR process. Nevertheless, users should be
aware that the newgeneral model in its default condition
does not result in a COD balance across the system. To eliminate
COD losses (and thus force a COD balance across a newgeneral
-based layout) each of these stoichiometric loss terms should be
set to 1.0.
Included in newgeneral
are two new heterotrophic yield terms to account for differences in
yield under varying electron acceptor conditions. Where the
general model assumed the same yield irrespective of the
electron acceptor, newgeneral differentiates the yield
depending on the presence of oxygen and nitrate. The added
stoichiometric terms include:
·
Yield (anoxic)
·
Yield (anaerobic)
In total, two state variables and eight
processes have been added to the general model.
State Variables
·
Non-releasable polyphosphate (XPP)
·
Soluble inorganic nitrogen (SNI)
Processes
·
Anoxic hydrolysis of stored/enmeshed COD
·
Anaerobic hydrolysis of stored/enmeshed COD
·
XPP lysis on aerobic decay
·
XPP lysis on anaerobic decay
·
Anoxic decay of PolyP organisms
·
XPPR lysis on anoxic decay
·
XPP lysis on anoxic decay
·
XBT lysis on anoxic decay
Phosphorus Uptake
As evidence exists
that not all polyphosphate is releasable, a new state variable to
represent non-releasable polyphosphate was incorporated. The uptake
of phosphorus (both aerobically and anoxically) in
newgeneral results in the storage of both releasable and
non-releasable polyphosphate. The partitioning of the storage is
stoichiometrically determined based on the total phosphorus taken
up.
PolyP Decay
In the
general model decay of PolyP organisms was modelled as an
aerobic or anaerobic process. The newgeneral model includes
anoxic decay of XBP to those processes. This results in the anoxic
lysis of XBT and XPPR modelled in a similar way to the aerobic
lysis of these variables in the general model. To account
for the lysis of XPP, newgeneral includes three new
processes: one for aerobic lysis, one for anoxic lysis and one for
anaerobic lysis.
Hydrolysis
The hydrolysis of
particulate COD to soluble material suitable for growth is a
critical step in the COD cycle. The general model modelled
hydrolysis as an aerobic process only, butnewgeneralincludes
hydrolysis under anoxic and anaerobic conditions and uses these
processes as a sink for COD 'loss.
Fermentation
Whereas the
fermentation of SS to SLF was a non-growth related process in the
general model, in newgeneral, this process is
associated with the growth of heterotrophic organisms and hence, a
portion of the fermented COD winds up as new biomass; however,
equally important is the stoichiometry of this process because,
like hydrolysis, this process includes a COD 'loss' component.
The Model Matrix for the
newgeneral model is found in Appendix A.
Introduction
A new comprehensive
model incorporating the most commonly observed biological,
physical, and chemical processes in wastewater treatment plants was
developed and implemented in GPS-X. MANTIS2 is the outcome of
Hydromantis’ commitment to provide state of the art models to its
clients through research and development. The MANTIS2 model
incorporates a large amount of information which has become
available in technical literature in the last decade. The
motivation behind the development of a comprehensive model arises
from the fact that after the publications of ASM2d and ADM1, there
has been growing interest in extending the capabilities of models
to incorporate the side stream treatment processes like struvite
precipitation, nitrification-anammox for nitrogen removal, and
other precipitation processes. Although state variable interfaces
between ASM2d and ADM1 are well defined, the state variable mapping
between two models present practical challenges in model
implementation and limits the versatility in modelling.
MANTIS2 Model
Components
The key features of
the MANTIS2 model are:
·
Carbon, Nitrogen, and Phosphorus removal with integrated anaerobic
digestion processes.
·
Mass balance for COD, C, N, P, Ca, Mg, K, and charge
·
48 state variables (21 soluble + 27 particulate)
·
56 processes
·
Two-step nitrification using AOB, NOB
·
Two-step denitrification
·
Methanol degradation process with methylotroph biomass
·
ANAMMOX process
·
Anaerobic digestion processes with gas phase modelling for
N2, CO2, H2, and
CH4
·
pH and alkalinity estimation in both liquid and solid train
·
Precipitation of MgNH4PO4. 6H2O,
CaCO3, MgHPO4. 3H2O,
CaPO4, AlPO4, FePO4,
Al(OH)3 and Fe(OH)3
·
Unified composite variable calculation
The MANTIS2 model is is implemented in MANTIS2LIB. The MANTIS2
model allows estimation of pH in both the liquid and solid train,
therefore it is possible in MANTIS2 to use additional influent
objects like acid feed, alkali feed to estimate the chemical
required for pH adjustment.
The basic structure of the comprehensive MANTIS2 model is based on
the following published models:
1.
ASM2d – basic reactions for biological carbon, nitrogen, and
phosphorus removal
2.
UCTADM1 – Anaerobic digestion processes
3.
Musvoto Model – Inorganic precipitation processes
In addition to the above, reference is made to other accepted
models (i.e. newgeneral model, ADM1 and UCTCN and MANTIS) in
formulating the model structure and finalizing the default
stoichiometry and kinetics of model parameters. The modelling
approaches for two-step nitrification, two-step denitrification,
denitrification on methanol and ANNAMOX are adopted from recent
research studies.
State Variables
A complete
description of the state variables in MANTIS2 model is located in
the Comprehensive Model Library (MANTIS2LIB) section of
0.
The soluble states
in the model can be classified into four categories: Soluble
inert organic; Soluble biodegradable organics;
Soluble inorganic ions; and Dissolved gases. The
states in each category are described in following sections.
Soluble Inert
Organics
1.
Soluble inert organics (si)
The state
represents the inert organics in the wastewater. The soluble
inert organics are not degraded in the wastewater treatment
processes. MANTIS2 model considers the generation of soluble inert
organic during the hydrolysis of slowly biodegradable substrate. By
default, the fraction of soluble inert organic production is set to
zero. In the anaerobic digestion process, the fraction may be set
to a non-zero value to calibrate the soluble inert COD from the
digesters.
Soluble
Biodegradable Organics
1.
acetate (sac)
2.
propionate (spro)
3.
methanol (smet)
4.
fermentable substrate (ss)
5.
colloidal substrate (scol)
The state of
fermented substrate (slf) in ASM2D is split into two states
of acetate (sac) and propionate (spro). The inclusion of an
additional state of spro is required as it is one of the
important intermediate products in anaerobic digestion.
The additional
state of methanol (smet) is added to the model. It is now
well accepted that the single carbon methanol is degraded by
special microorganism which have different kinetics than the
ordinary heterotrophic organism. This additional state is helpful
to track the degradation of methanol based on the degradation
kinetics of methylotrophic biomass.
Although there is
no well accepted definition of the state of colloidal substrate in
wastewater, for the purpose of modelling, it is assumed that this
fraction represents the portion of the substrate COD which lies in
the size range of 0.45 micrometers to 1.2 micrometers. These are
typically macromolecules which require hydrolysis before oxidation.
This fraction of COD is assumed to behave differently in different
unit processes for example the colloidal COD will behave:
·
as soluble in solid liquid separation in settlers
·
as particulate in biological degradation
·
as particulate or soluble in membrane separation
Soluble
Inorganic
In addition to the
soluble organic states, the MANTIS2 model considers inorganic
states of soluble inorganic carbon (stic), soluble nitrite-N
(snoi), soluble nitrate-N (snoa), soluble Ammonia-N
(snh), soluble Organic nitrogen (snd), soluble
ortho-P (sp), dissolved calcium (sca), dissolved
(smg), dissolved potassium (spot), dissolved anion
(sana) and dissolved cation (scat). These inorganic
states are chosen so as to appropriately describe the N and P
transformation, inorganic precipitation and pH changes across the
various unit processes in wastewater treatment plant.
The soluble
inorganic carbon, stic, represents the sum of carbon in all
the ionic species in the carbonic acid system i.e.
H2CO3, HCO3-,
CO32-. In previous models like ASM2D and New
General, the alkalinity (salk) is used to express the buffer
capacity in the wastewater. In MANTIS2, soluble inorganic carbon is
used instead of alkalinity as the state variable as it is a
conserved quantity. The alkalinity in the model is estimated by
considering the soluble inorganic carbon and the estimated pH of
the system.
Two oxidized form
of soluble nitrogen, (i.e. soluble nitrite and nitrate) are
considered in the model. This choice is necessary to model the two
step nitrification process.
The additional
inorganic states like soluble calcium (sca), soluble
magnesium (smg) are introduced to model the key
precipitation reaction involving these ionic species. The state
variable of soluble potassium (spot) is included to model
the uptake and release of potassium during polyphosphate formation
and degradation. The dissolved anion (sana) and cation
(scat) represent all other strong anion and cation in the
wastewater. These states are used in formulating the charge balance
equation for estimating the pH in the wastewater.
Soluble
Gases
In addition to
soluble oxygen (so), MANTIS2 includes soluble gases states
for nitrogen (sn2), methane (sch4) and hydrogen
(sh2). The soluble CO2 is estimated by the pH and
the stic concentration in the wastewater. The soluble oxygen
plays an important role in the aerobic biological systems. The
other gases are more relevant in the anaerobic digestion and
fermentation processes. For each soluble gas, gas-liquid transfer
equation are used to model the dissolution/stripping of the gas in
the unit processes.
The particulate
states in the model are classified into four categories of
Particulate Inert Organics, Particulate Organic
Substrate/Storage, Active Biomass and Particulate
Inorganic.
Particulate
Inert Organics
The model includes
two states of Inert Organic Particulate (xi) and
Unbiodegradable Organic Matter from cell decay (xu) in the
model. The inert organic particulate is the inert organics fraction
that is contributed by the influent wastewater. Having two states
for inert organic compounds makes it easier to differentiate
between the amount of inert organics accumulated from the
wastewater and the amount that is produced by cell decay in the
system.
Particulate
Organic Substrate/Storage
MANTIS2 provides
one particulate Organic Substrate (xs) and one organic
storage compound for intracellular PHA (xbt). These states
are equivalent to the states in ASM2D and New General Models.
Active
Biomass
Several biomass
states are added in MANTIS2. To model the two step nitrification
process, two autotrophic biomass Ammonia Oxidizer (xbai) and
Nitrite Oxidizer (xbaa) are considered. Methylotrophic
biomass (xbmet) is included to model the biodegradation of
methanol. The fermentive biomass (xbf), Acetogen
(xbpro), acetate methanogens (xbacm) and hydrogen
methanogens (xbh2m) are included to model the anaerobic
transformation in anaerobic digestion.
Particulate
Inorganic
The particulate
inorganic states include the model precipitates like aluminum
hydroxide, aluminum phosphate, iron hydroxide, iron phosphate,
calcium carbonate, calcium phosphate, magnesium hydrogen phosphate,
magnesium carbonate, and ammonium magnesium phosphate (struvite).
In addition to these precipitates, particulate inert inorganic is
used as a composite state for the unidentified inorganic in the
wastewater. The N and P components associated with the slowly
biodegradable organics are included as nitrogen in slowly deg.
organics and phosphorous in slowly deg. organics. The inorganic
poly-phosphate accumulated in PAO is also included as a state.
Composite Variables
MANTIS2 uses a
well-defined stoichiometry for each state variable to estimate the
composite variables. The state variables and their relationships to
the composite variables are shown in the Composite Variables in
MANTIS2LIB section of 0, GPS-X Composite
Variable Calculations.
Processes
The processes
included in the MANTIS2 model are described below:
Adsorption/Enmeshment
1.
Adsorption of colloidal COD: The colloidal COD
(scol) is considered to first adsorb on the heterotrophic
biomass. The adsorbed colloidal COD then becomes a part of slowly
biodegradable COD (xs) and requires hydrolyses before it
become available for bacterial metabolism. As both the ordinary
heterotrophic organism and fermentative organism are considered to
participate in the hydrolysis process, the rate of adsorption is
defined with respect to the sum of the concentration of two
organisms. The adsorption rate is first order to the colloidal COD.
A rate inhibition term is added to reduce the rate of adsorption as
the ratio of slowly biodegradable COD to adsorbing biomass
increases.
Processes
mediated by heterotrophic organisms
2.
Aerobic hydrolysis: The heterotrophic
microorganisms are considered to participate in the hydrolysis of
slowly biodegradable substrate XS resulting in
production of soluble fermentable substrate (ss). Surface
limited hydrolysis kinetics similar to that used in ASM1 and ASM2d
is used. Both the ordinary heterotrophic organism and fermentative
organism are considered to participate in the hydrolysis
process
3.
Anoxic hydrolysis: This hydrolysis process is
active under anoxic conditions. The oxygen saturation term in
aerobic hydrolysis rate expression is replaced with an oxygen
inhibition term. A NOX
(NO2-N+NO3-N) saturation term is added to the
rate expression. The specific hydrolysis rate is reduced by anoxic
hydrolysis reduction factor (ηnox).
4.
Anaerobic hydrolysis: This hydrolysis process
is active only under anaerobic conditions. The rate expression
contains oxygen and NOX inhibition terms. The specific
hydrolysis rate is reduced by anaerobic hydrolysis reduction factor
(ηanaer).
5.
Ammonification: The ammonification process
converts soluble organic nitrogen to ammonia nitrogen. Both the
ordinary heterotrophic organism and fermentative organism are
considered to participate in the ammonification process. The
kinetics of ammonification is similar to that provided in ASM1.
6.
Growth on fermentable substrate (ss) using O2 as
electron acceptor: The process of heterotrophic
growth takes place under aerobic conditions. The reaction rate for
this process is formulated considering the concept of multi
substrate kinetics outlined in ASM2d model. The growth rate is
considered proportional to the ratio of fermentable substrate to
total soluble substrate (ss+sac+spro)
available to heterotrophic biomass. The main difference in the
growth stoichiometry of different biomass is the uptake of N and P,
Ca, Mg, K, anion and cation during the biomass growth. In this
implementation, it is assumed that concentration of Ca, Mg and K is
non-limiting during growth. If required, the saturation terms for
each micronutrient can be added easily in the model
equations. The growth kinetics does not use the alkalinity
saturation function as it is planned to add a pH inhibition term in
the growth kinetics at a later date.
7.
Growth on acetate (sac) using O2 as electron
acceptor: The process of aerobic heterotrophic growth
on acetate is similar to the process #6, except that the growth
rate is proportional to the ratio of acetate concentration to the
total soluble substrate (ss+sac+spro) available to
heterotrophic biomass.
8.
Growth on propionate (spro) using O2 as electron
acceptor: The process of aerobic heterotrophic growth
on acetate is similar to the process #6, except that the growth
rate is proportional to the ratio of propionate concentration to
the total soluble substrate (ss+sac+spro)
available to heterotrophic biomass.
9.
Growth on fermentable substrate (ss) using NO3 as
electron acceptor: This process of heterotrophic
growth takes place in the presence of NO3-N. The
stoichiometry of this process is developed by considering partial
reduction of NO3-N to NO2-N. The rate
expression for the process uses an inhibition term for oxygen and a
saturation term for the NO3-N. The growth rate is also
considered proportional to the ratio of fermentable substrate to
total soluble substrate (ss+sac+spro) available to
heterotrophic biomass. The reaction rate expression also reduces
the amount of heterotrophic biomass by the fraction of
NO3-N to total NOX nitrogen available in the
system. It is assumed that when the heterotrophic biomass is
converting NO3-N to NO2-N, it is not
participating in the conversion of NO2-N to
N2 gas.
10.
Growth on acetate (sac) using NO3 as
electron acceptor: The process of heterotrophic
growth on acetate using NO3-N is similar to the process
#9, except that the growth rate is proportional to the ratio of
acetate concentration to the total soluble substrate
(ss+sac+spro) available to heterotrophic
biomass.
11.
Growth on propionate (spro) using NO3 as electron
acceptor: The process of heterotrophic growth on
acetate using NO3-N is similar to the process #9, except
that the growth rate is proportional to the ratio of propionate
concentration to the total soluble substrate (ss+sac+spro)
available to heterotrophic biomass.
12.
Growth on fermentable substrate (ss) using
NO2 as electron acceptor: This process of
heterotrophic growth takes place in the presence of
NO2-N. The stoichiometry of this process is developed by
considering reduction of NO2-N to N2.
The rate expression for the process uses an inhibition term for
oxygen and a saturation term for the NO2-N. The growth
rate is also considered proportional to the ratio of fermentable
substrate to total soluble substrate (ss+sac+spro) available
to heterotrophic biomass. The reaction rate expression also reduces
the amount of heterotrophic biomass by the fraction of
NO2-N to total NOX nitrogen available in the
system. It is assumed that when the heterotrophic biomass is
converting NO2-N to N2, it is not participating in the
conversion of NO3-N to NO2-N gas.
13.
Growth on acetate (sac) using NO2 as
electron acceptor: The process of heterotrophic
growth on acetate using NO2-N is similar to the process
#12, except that the growth rate is proportional to the ratio of
acetate concentration to the total soluble substrate
(ss+sac+spro) available to heterotrophic biomass.
14.
Growth on propionate (spro) using NO2 as electron
acceptor: The process of heterotrophic growth on
acetate using NO2-N is similar to the process #12 except
that the growth rate is proportional to the ratio of propionate
concentration to the total soluble substrate (ss+sac+spro)
available to heterotrophic biomass.
15.
Decay of heterotrophs: The process rate of
heterotrophic decay is modeled similar to that in ASM2d model. The
main difference in the stoichiometry of decay reaction is the
production of N, P, Ca, Mg, K, anion and cation according to the
biomass composition
Processes
mediated by autotrophic organisms
The model considers
two-step conversion of NH3-N to NO3-N. The
two steps are mediated by ammonia oxidizer and nitrite oxidizer
sequentially.
16.
Growth of ammonia oxidizer: The process of
growth of ammonia oxidizer oxidizes NH3-N to NO2-N in the presence
of oxygen. The reaction rate uses ammonia and oxygen saturation
terms. The stoichiometry of the process is developed based on the
conversion of NH3-N to NO2-N.
17.
Growth of nitrite oxidizer: The process of
growth of ammonia oxidizer oxidizes NO2-N to
NO3-N in the presence of oxygen. The reaction rate uses
NO2-N and oxygen saturation terms. The stoichiometry of
the process is developed based on the conversion of
NO2-N to NO3-N.
18.
Decay of ammonia oxidizer: The process rate of
ammonia oxidizer decay is modelled similar to that in the ASM2d
model.
19. Decay of
nitrite oxidizer: The process rate of nitrite
oxidizer decay is modelled similar to that in the ASM2d model.
Processes
mediated by phosphate accumulating organisms (PAO)
The processes
mediated by PAO are based on the ASM2D model. Three new processes
describing the storage of PHA on propionate, growth of PAO on PHA
using NO2 as electron acceptor and anoxic storage of XPP
using NO2 are added to the processes mediated by
PAO.
20. Storage of
PHA by PAO using acetate: The rate expression for PHA
storage by PAO using acetate as used in ASM2d is modified to
include the effect of propionate, another VFA used by PAO. The rate
expression is modified by assuming that PAO can utilize acetate and
propionate simultaneous in proportion to the availability of each
substrate
21.
Storage of PHA by PAO using propionate: This is
a new process and describes the storage of PHA by PAO using
propionate. The kinetics and stoichiometry of the process is based
on concepts used in describing process #20.
22.
Growth of PAO on PHA using O2 as electron
acceptor: The stoichiometry and rate expression for
this process is the same as ASM2d.
23.
Storage of XPP on PHA using O2 as
electron acceptor: The stoichiometry and rate
expression for this process is the same as ASM2d.
24.
Growth of PAO on PHA using NO3 as electron
acceptor: This process of PAO growth takes place in
the presence of NO3-N. The stoichiometry of this process
is developed by considering partial reduction of NO3-N
to NO2-N. The rate expression for the process uses
an inhibition term for oxygen and a saturation term for the
NO3-N. The reaction rate expression also reduces the
amount of PAO biomass by the fraction of NO3-N to total
NOX nitrogen available in the system. It is assumed that
when the PAO biomass is converting NO3-N to
NO2-N, it is not participating in the conversion of
NO2-N to N2 gas.
25.
Storage of XPP on PHA using NO3 as
electron acceptor: In this process the storage of
poly-phosphate takes place while the stored PHA compounds are
oxidised using NO3-N as an electron acceptor. Only
partial reduction of NO3-N to NO2-N is
considered in process stoichiometry formulation. Similar to the
process #24, a reduction factor equal to the ratio of
NO3-N to NOX-N concentration is applied to
account for fraction of total biomass mediating this process.
26.
Growth of PAO on PHA using NO2 as electron
acceptor: This process of PAO growth takes place in
the presence of NO2-N. The stoichiometry of this process
is developed by considering partial reduction of NO2-N
to N2-N. The rate expression for the process uses
an inhibition term for oxygen and a saturation term for the
NO2-N. The reaction rate expression also reduces the
amount of PAO biomass by the fraction of NO2-N to total
NOX nitrogen available in the system. It is assumed that
when the PAO biomass is converting NO2-N to
N2-N, it is not participating in the conversion of
NO3-N to NO2-N.
27.
Storage of XPP on PHA using NO2 as
electron acceptor: In this process the storage of
poly-phosphate takes place while the stored PHA compounds are
oxidised using NO2-N as an electron acceptor. The
conversion of NO2-N to N2 is considered in
process stoichiometry formulation. Similar to the process #26, a
reduction factor equal to the ratio of NO2-N to
NOX –N concentration is applied to account for fraction
of total biomass mediating this process.
28.
Decay of PAO: The stoichiometry and rate
expression for this process is same as ASM2d, except that
alkalinity saturation term is not used.
29.
XPP lysis: The stoichiometry and rate
expression for this process is same as ASM2d, except that
alkalinity saturation term is not used.
30. PHA
lysis: The stoichiometry and rate expression for this
process is same as ASM2d, except that alkalinity saturation term is
not used.
Processes
mediated by methylotrophs
A single population
of methylotrophs, which degrades methanol, a single carbon
substrate is incorporated in the model. Four reaction
processes are considered for this biomass.
31.
Growth of methylotrophs on methanol using O2 as
electron acceptor: The stoichiometry and kinetics of
the process is developed using the principles of heterotrophic
growth. The rate expression for growth of methylotrophs includes
oxygen saturation and a methanol saturation term. The biomass yield
on methanol is reported to be much lower than other carbon sources.
Therefore, a different yield coefficient is used in the process
stoichiometry.
32. Growth of
methylotrophs on methanol using NO3 as electron
acceptor: The methylotrophs can use NO3-N
as terminal electron acceptor to oxidize methanol. The
stoichiometry of this process is developed by considering partial
reduction of NO3-N to NO2-N. The rate
expression for the process uses an inhibition term for oxygen and a
saturation term for the NO3-N. The reaction rate
expression also reduces the amount of biomass mediating the
reaction by the fraction of NO3-N to total
NOX nitrogen available in the system. It is assumed that
when the methylotroph biomass is converting NO3-N
to NO2-N, it is not participating in the conversion of
NO2-N to N2 gas.
33.
Growth of methylotrophs on methanol using NO2 as
electron acceptor: This represents the second step of
denitrification. The stoichiometry of this process is developed by
considering partial reduction of NO2-N to
N2-N. The rate expression for the process uses an
inhibition term for oxygen and a saturation term for the
NO2-N. The reaction rate expression also reduces the
amount of biomass mediating the reaction by the fraction of
NO2-N to total NOX nitrogen available in the
system. It is assumed that when the methylotroph biomass is
converting NO2-N to N2-N, it is not
participating in the conversion of NO3-N to
NO2-N.
34. Decay of
methylotrophs: Decay of methylotrophs is modelled
using first order reaction rate, with respect to the biomass
concentration.
Processes
mediated by Anaerobic Microorganisms
The MANTIS2 model
includes processes mediated by anaerobic microorganisms. These
processes present the key transformations observed in strict
anaerobic environment like anaerobic digester and other
modifications of anaerobic treatment technology. These
processes are adapted from the anaerobic digestion model developed
at University of Cape Town (UCTADM1). The process rate and
stoichiometry was converted from molar units to COD units. The
process stoichiometry was also modified to include the mass
balances for phosphorus, Ca, Mg, K, cation and anion species. The
scheme of anaerobic biodegradation assumes that the anaerobic
biomass decay takes place according to the decay processes listed
above. The slowly degradable substrate is then hydrolyzed
anaerobically according to process 4. The resulting soluble
fermentable substrate is then converted to CH4 and
H2 by the processes described below.
35.
Growth of fermentive bacteria at low
H2: This process models the fermentation
by acidogens under low H2 partial pressure. The process
stoichiometry is developed based on a conversion of model
fermentable substrate (glucose) to acetic acid, H2 and
CO2. The kinetic expression for the process uses a
H2 inhibition term to reduce the reaction rate as the
partial pressure of H2 increases.
36. Growth of
fermentive bacteria at high H2: This
process models the fermentation by acidogens under high
H2 partial pressure. The process stoichiometry is
developed based on a conversion of model fermentable substrate
(glucose) to acetic acid, propionic acid, H2 and
CO2. The kinetic expression for the process uses a
H2 saturation term to account for increased rate at
higher partial pressure of H2
37.
Decay of fermentive biomass: The stoichiometry
and kinetic expression for the decay process of fermentative
biomass is based on the same principles as for other biomass
types.
38.
Growth of acetogens on propionate: This process
(Acetogenesis) is the process whereby the acetogens convert
propionic acid under low hydrogen partial pressure. The kinetic
rate expression for the process includes a H2 inhibition
term and a propionic acid saturation term.
39.
Decay of acetogens: The stoichiometry and
kinetic expression for the decay process of fermentative biomass is
based on the same principles as for other biomass types.
40.
Growth of hydrogenotrophic methanogens: In this
process hydrogenotrophic methanogens grow by converting
H2 and CO2 to CH4. The
stoichiometry of the process is based on the chemical reaction
describing this process. The kinetic expression for the process
uses a hydrogen saturation term.
41.
Decay of hydrogenotrophic methanogens: The
stoichiometry and kinetic expression for the decay process of
fermentative biomass is based on the same principles as for other
biomass types.
42.
Growth of Acetoclastic methonegens: In this process
the the acetic acid is converted to methane and CO2 by
acetoclastic methanogens. The kinetic expression of the process
includes an acetate saturation term.
43.
Decay of acetoclastic methanogens: The stoichiometry
and kinetic expression for the decay process of fermentative
biomass is based on the same principles as for other biomass
types.
Processes
mediated by anaerobic autotrophic microorganisms
A new type of anaerobic autotrophic biomass type is included in
MANTIS2 to perform anaerobic ammonium oxidation (ANAMMOX) process.
The process stoichiometry is described in Strous et al. (1998) as
below.
1NH4+ + 1.32 NO2- +
0.066 HCO3- + 0.13 H+
à 1.02 N2
+ 0.26 NO3- + 0.066
CH2O0.5N0.15 +2.03
H2O
According to the
above reaction, a mole of ammonia-N is oxidized to 1.02 mole of
N2 using 1.32 mole of nitrite-N. The reaction also
produces 0.26 mole of nitrate-N and leads to growth of autotrophic
biomass.
To model this
process, two processes are added in MANTIS2.
44.
Growth of anammox microorganism: The
stoichiometry factors to express the yield of biomass, NO2-N
consumption and N2 and NO3- are expressed in terms of per unit
NH4+-N oxidized. The reaction rate
expression uses an oxygen inhibition function and NO2-N
and NH3-N saturation functions. Total (ionized +
non-ionized) concentration of both substrates is used in the
kinetic expression.
45. Decay of
anammox microorganism: Decay of anammox
microorganisms is modelled using first order reaction rate with
respect to the biomass concentration.
Chemical
precipitation processes
The precipitation
reactions in MANTSI2 are adapted from Musvoto et
al. 2000 with some modifications. In addition to the
processes of precipitation of AlPO4 and FePO4
which are included in ASM2d, five commonly observed precipitation
processes in wastewater treatment are included in the model. The
key differences from
Musvoto et al. 2000 are: 1) the
effect of ion-pair is neglected in the calculations; 2) the kinetic
rate equations use the approach suggested in ASM2d with a few
modification and 3) the mole based stoichiometry is replaced with
mass based stoichiometry for maintaining model consistency. The
kinetic expression for the precipitation reactions uses the
concentration of participating ions, rate of precipitation and
solubility product of the precipitate. The kinetic rate expression
is formulated such that the precipitation and dissolution reactions
can be modeled by a single equation.
46.
Precipitation/dissolution of CaCO3:
CaCO3 is assumed to precipitate as calcite. The calcite
precipitation is assumed to take place in the presence of
Ca2+ and CO32- ions in the
solution. The concentration of CO32- is
estimated by using the soluble inorganic carbon and the pH of the
system.
47.
Precipitation/dissolution of
MgNH4PO4.6H2O
(struvite): Struvite is the most commonly observed
precipitate in the digester supernatant. The precipitation of
struvite takes place if the Mg2+,
NH4+ and PO43- species
are present in the solution. The concentrations of
NH4+ and PO43- are
estimated based on the solution pH.
48.
Precipitation/dissolution of
MgHPO4.3H2O (newberyite): This
is another precipitate of magnesium with phosphate which is
normally observed at lower pH. For this precipitate,
Mg2+ and HPO42- ionic species are
required. The concentration of HPO42- ion is
estimated based on pH of the solution.
49.
Precipitation/dissolution of Ca3(PO4)2
(ACP): As the numbers of water molecules associated with
the precipitate are variable, an anhydrous form is considered in
expressing the mass concentration of the precipitate. It is
indicated that this is probably the least stable precipitate among
the possible precipitates and transforms into more stable forms
with time. The Ca2+ ion and PO43-
species are required in precipitation process.
50.
Precipitation/dissolution of MgCO3:
MgCO3 (magnesite) is another precipitate of
magnesium that is included in the model. The precipitation reaction
requires Mg2+ and CO32- ionic
species for its formation.
51.
Precipitation/dissolution of AlPO4:
The precipitation of AlPO4 is required to model the
metal precipitation of phosphorous in the plant. The stoichiometry
of the process is similar to ASM2d; however, the kinetic expression
is modified to include the solubility product of the
precipitate.
52.
Precipitation/dissolution of FePO4:
The precipitation of FePO4 is required to model the
metal precipitation of phosphorous in the plant. The stoichiometry
of the process is similar to ASM2d; however, the kinetic expression
is modified to include the solubility product of the
precipitate.
Gas liquid
transfer processes
Similar to the gas
liquid transfer of oxygen, four additional processes as below are
added in the model. The gas-liquid transfer is modeled using a mass
transfer constant (KLa) and the gas saturation
concentration at the given temperature and pressure. The mass
transfer constant for each gas can be correlated to the oxygen mass
transfer using gas diffusivities. For simplification, fractional
factors are applied to the mass transfer constant of oxygen to
obtain respective mass transfer constant.
53.
Gas liquid transfer of CO2: Process
describes the stripping/absorption of CO2 from/to the
liquid. Since the concentration of CO2 in liquid depends
on the pH, this process is very sensitive to the pH of the
solution.
54.
Gas-liquid transfer of N2: Process
describes the stripping/absorption of N2 from/to the
liquid.
55.
Gas-liquid transfer of CH4: Process
describes the stripping/absorption of CH4 from/to the
liquid.
56.
Gas-liquid transfer of H2: Process
describes the stripping/absorption of H2 from/to the
liquid.
Algebraic pH Solver
In the MANTIS2 model, the pH is estimated in all the unit
processes. The dissociation reactions for acid and bases are much
faster than the other processes used in the model, therefore,
algebraic form of equations are chosen over the differential form
for expressing acid/base dissociation reactions. The pH in each
unit process is estimated by solving a set of algebraic equations
which include a charge balance equation along with the equilibrium
equations for each ionic species. Although temperature dependency
of the dissociation constant is considered in the model, ionic
activity corrections are not applied in the equations. For a
typical wastewater, the ionic activity correction may not be
important, but at higher ionic concentrations the model pH values
should be used with appropriate caution.
The MANTIS2 model
uses the dissociation equations for: carbonic acid
(diprotonic), phosphoric acid (triprotonic), ammonium
(monoprotonic), acetic acid (monoprotonic), propionic
acid (monoprotonic), nitrous acid (monoprotonic). The
charge balance equation to solve for [H+] concentration
is prepared using the ionized species of weak acid/base and strong
anions (NO3-, sana) and strong cations
(Ca2+, Mg2+, K+, scat).
The
pre-fermentation model implemented in GPS-X is based on the
Munch et al. model (1999), which is a
mechanistic model aiming at describing the effect of design and
operating parameters on the rate of VFAs production. The
pre-fermenter model is found in the CSTR object only.
A general reaction
pathway is shown in Figure 6‑11.
Figure 6‑11 - General Reaction Pathway
The state variables
in the prefermenter model are as follows:
·
Cis: insoluble substrate (not passing
through a 0.45 mm filter) such as cellulose, fats, insoluble
proteins (mg COD/L).
·
Css: soluble, high molecular weight
substrate, such as soluble carbohydrates (starch), and soluble
proteins (globular) (mg COD/L).
·
Cmo: monomer species such as glucose,
long-chain fatty acids, amino acids (mg COD/L).
·
Slf: volatile fatty acids (mg COD/L).
·
Ce: hydrolytic enzymes which act as a catalyst to
hydrolyse insoluble substrate, soluble high-molecular-weight
substrate and proteins (mg COD/L).
·
Cxa: acidogenic bacteria which utilize
monomers (Cmo) as their substrate (mg COD/L).
·
Cxm: methanogenic bacteria, which utilize
volatile fatty acids (VFAs) as their substrate then transforming it
into methane gas (mg COD/L).
·
Cprot: organic nitrogen contained in
particulate proteins which is converted into ammonia nitrogen in
the ammonification process as particulate proteins are hydrolysed.
It is also produced from the decay of bacteria cells (mg N/L).
·
Snh: ammonia nitrogen (mg N/L).
·
CH4,g: methane gas.
Stoichiometric
Parameters
The stoichiometric
parameters to be specified in the influent section are used to
calculate the concentration of soluble, high molecular weight
substrate (Cis) from the influent readily biodegradable
substrate (ss) and the concentration of methanogenic biomass
(Cxm) from the influent slowly biodegradable substrate
(Xs):
·
fss: fraction of soluble high molecular weight
substrate (Cis) in the influent readily biodegradable
substrate (ss).
·
fxm: fraction of methanogenic biomass
(Cxm) in the influent slowly biodegradable substrate
(Xs).
·
fm: conversion factor from COD to cubic meters
for methane production (m3 CH4/g
COD).
The stoichiometric
parameters specified in the “hydrolysis”, “acidogens” and
“methanogens” sections are the following yield coefficients:
·
Ye: Yield for hydrolytic enzymes on
insoluble or soluble substrate (Ce/Css).
·
Ya: yield for acidogens on monomers species
(Cxa/Cmo).
·
Ym: Yield for methanogens on volatile fatty acids
(Cxm/Slf).
Kinetic Parameters
The model kinetic
parameters are:
·
khis: hydrolysis rate constant for insoluble
substrate (g COD/m3.d).
·
khs: hydrolysis rate constant for soluble
substrate (g COD/m3.d).
·
kamm: ammonification rate constant for organic
nitrogen contained in proteins (g N/m3.d).
·
kmo: maximum specific consumption rate of
monomers by acidogenic biomass (g COD/m3.d).
·
kac: maximum specific consumption rate of
volatile fatty acids by methanogenic biomass (g
COD/m3.d).
·
da: decay rate constant of acidogens (g
COD/m3.d).
·
dm: decay rate constant of methanogens (g
COD/m3.d).
·
de : deactivation rate constant of hydrolytic
enzymes (g COD/m3.d).
Process Rates
The process rates
described in the model are:
Hydrolysis of
insoluble substrate
Equation 6.45
The hydrolysis rate
of insoluble substrate is considered to be first order with respect
to the concentration of insoluble substrate and to the
concentration of hydrolytic enzymes, and to be inversely
proportional to the concentration of acidogenic biomass, due to a
limited surface area available which could cause mass transfer
limitations.
Hydrolysis of
soluble substrate
Equation 6.46
The hydrolysis of
soluble substrate is considered to be first order with respect to
the concentration of soluble substrate and to the concentration of
hydrolytic enzymes.
Ammonification
of proteins
Equation 6.47
The ammonification
of proteins is considered to be proportional to the concentration
of proteins, to the concentration of hydrolytic enzymes and
inversely proportional to the acidogenic biomass
concentration. Proteins are regarded as particulate matter
with surface area limitations.
Consumption of
monomers by acidogens
Equation 6.48
The consumption
rate of monomers is assumed to be proportional to monomers and to
ammonia nitrogen saturation functions and to the concentration of
acidogenic biomass.
Consumption of
VFAs by methanogens
Equation 6.49
The consumption
rate of volatile fatty acids is considered to be proportional to
the volatile fatty acids and ammonia nitrogen saturation functions
and to the concentration of methanogenic biomass.
Decay of
acidogens
Equation 6.50
The rate of decay
of acidogens is assumed to be proportional to their
concentration.
Decay of
methanogens
Equation 6.51
The rate of decay
of methanogens is assumed to be proportional to their
concentration.
Decay of
enzymes
Equation 6.52
The rate of decay of enzymes is assumed to be proportional to their
concentration
Model Structure
A schematic diagram
of the prefermenter model is shown in Figure
6‑12 (adapted from Munch et al.,
1999).
Figure 6‑12 – Schematic Diagram of the
prefermenter Model
The Model Matrix
for the prefermenter model can be found in Appendix A.
With a large number
of models available in the provided literature and used in GPS-X,
it can be a challenge to select which model is best for each
modelling application. The following simple guide gives a few
suggestions and “rules of thumb” for selecting an activated sludge
model.
1.
The first choice to be made is the GPS-X macro library. If you are
only concerned with carbon and nitrogen processes, then you should
use the CN (1‑step nitrification) library. If you are interested in
modelling phosphorus, then the CNP or Comprehensive libraries
should be used. If you are interested in modelling pH, inorganic
precipitation and/or side stream processes, use the Comprehensive
library. Extra user-defined components can be added to the
models by selecting the associated IP (Industrial Pollutant)
library – CNIP or CNPIP.
2.
It is a good rule of practice to keep the model as simple as
possible at the beginning, until you interpret the results and are
comfortable with that model's level of complexity. You can then
move to more complicated models. If you are unsure of which model
to start with we recommend using asm1, or mantis.
3.
The choice of model should depend on the amount and type of data
that is available to support its use. For example, if you have
little information about different substrate types in your system,
it is advisable to use models that have fewer substrates.
4.
If you are modelling a plant that has recycling of flows back from
the solids handling process back to the activated sludge line, the
Mantis2 model is best suited for handling this situation. The
Comprehensive (Mantis2) library contains a full set of state
variables that cover both activated sludge and anaerobic digestion
processes.
The choice of model
should reflect the need to simulate certain processes. For example,
if you are interested in exploring P-removal or alkalinity control,
you will need to use models that contain processes that are
relevant to those components. Table 6‑2 summarizes
the processes in each model, and may be useful in choosing a model.
You can consult the model descriptions earlier in this chapter or
the Model matrices found in Appendix A to see the details of the
processes contained in each model.
Table 6‑2 –
Model Processes in GPS-X
Process
|
asm1
|
asm3
|
mantis
(and 3dmantis)
|
asm2d
|
new general
|
mantis2
|
mantis3
|
Fermentation Step
|
|
|
|
X
|
X
|
X
|
X
|
Nitrification/Denitrification
|
X
|
X
|
X
|
X
|
X
|
X
|
X
|
Aerobic Denitrification
|
|
|
X
|
|
|
|
|
Aerobic Substrate Storage
|
|
X
|
|
|
|
|
|
COD "Loss"
|
|
|
|
|
X
|
|
|
2-Step
Nitrification/denitrification
|
|
|
|
|
|
X
|
X
|
NO3- as
a N source for cell synthesis
|
|
|
X
|
|
X
|
|
|
Alkalinity
consumption/generation
|
X
|
X
|
X
|
X
|
|
X
|
X
|
Alkalinity (as a limiting
factor for growth processes)
|
|
|
|
X
|
|
|
|
Biological phosphorus
removal
|
|
|
|
X
|
X
|
X
|
X
|
Precipitation of P with metal
hydroxides
|
|
|
|
X
|
|
X
|
X
|
Temperature dependency
|
X*
|
X*
|
X
|
X
|
X
|
X
|
X
|
pH
|
|
|
|
|
|
X
|
X
|
Struvite, other Calcium and
Magnesium ppt.
|
|
|
|
|
|
X
|
X
|
Anammox
|
|
|
|
|
|
X
|
X
|
Methylotroph
|
|
|
|
|
|
X
|
X
|
N2O gas production
|
|
|
|
|
|
|
X
|
*not part of the published model, but added in GPS-X.
The model
associated with the deep shaft object is an adaptation of the
mantis suspended-growth model. It takes into account the
specific hydraulics and hydrostatic pressure found in a deep shaft
reactor.
The deep shaft
header tank is modelled with two complete-mix reactors (by
default), and is under atmospheric pressure. The downcomer
and the riser shafts are modelled with twenty complete-mix reactors
(by default).
The two membrane
bioreactor objects in GPS-X (plug-flow and completely-mixed) are a
combination of a suspended-growth activated sludge model, and a
simple suspended solids separation filter. They are available
in all libraries, and with all biological models. The two MBR
objects are shown in Figure 6‑13, with the connection
points illustrated.
Figure 6‑13 - GPS-X Membrane Bioreactor
Objects
The structure of
the MBR model combines a conventional activated sludge tank model
(plug-flow or CSTR) with an in-tank solids separation filter, as
shown in Figure 6‑14. In the case of the
plug-flow MBR, the filter is placed in the final tank (an optional
internal recycle is shown for illustrative purposes).
Permeate flow is drawn through the filter at a rate determined by
the filter model.
Figure 6‑14 - Membrane Bioreactor Model
Structures
MBR Model Modes
The MBR models in
GPS-X allow for three modes of operation. The mode of operation can
be toggled in the Input Parameters > Model Options menu.
The three modes can be set up using the Model Option settings
summarized in Table 6‑3.
Table 6‑3 - Settings
of MBR Operating Modes
Operation Mode
|
Backwash and TMP calculations
|
Volume Calculations
|
Simple
|
Do not calculate backwash or
TMP
|
Fixed Volume
|
Intermediate
|
Calculate backwash and
TMP
|
Fixed Volume
|
Advanced
|
Calculate backwash and
TMP
|
Variable Volume
|
Simple Mode
assumes that the filter is properly operated to maintain flux, and
does not consider the effects of trans-membrane pressure (TMP),
cake formation, fouling, backwashing, and membrane resistance. The
Intermediate model mode calculates the TMP, cake formation,
fouling, backwashing, and membrane resistance based on the
specified backwash rate and cleaning frequency, but the required
permeate flux is calculated based on the influent flow and waste
flow to ensure that the liquid volume in the last tank remains
constant. The Advanced model mode is similar to the
Intermediate model mode, however the liquid volume in the tank is
variable and can increase or decrease based on the calculated
permeate flux. The user can control the liquid level using a
feedback controller that manipulates the trans-membrane pressure.
Table 6‑4 summarizes the differences between
Simple Mode, Intermediate, and Advanced
Mode.
Table 6‑4 –
MBR Model Modes
Operating Parameter
|
Operating Mode
|
Simple
|
Intermediate
|
Advanced
|
Flow Balance and Reactor Volume
|
Model assumes
flow in equals flow out, and there is no change in reactor
volume. All incoming flow is assumed to exit via the filter
and waste flow stream
|
Required
membrane flux is determined based on influent flow and waste flow
to ensure the liquid volume in the reactor is constant
|
Membrane flux
is determined from filter model. Reactor volume increases and
decreases depending on the difference between flow in and flow
out. A controller is provided to manage the tank level
|
Filter Operation
|
Filter
operation is ignored
|
Users must
specify TMP, backwash/relaxation cycles, and cross-flow
aeration
|
same as
intermediate
|
Cross-Flow Air
|
Filter-cake
solids removal from cross‑flow aeration is ignored, but oxygen
transfer from the cross‑flow air to the bulk liquid is calculated
and included in the biological activity
|
Both solids
removal and oxygen transfer are considered for cross-flow
aeration
|
same as
intermediate
|
Solids Capture
|
The solids
capture rate determines what fraction of the mixed liquor solids
remain in the reactor. These solids remain suspended in the
bulk liquid.
|
The solids
capture rate determines what fraction of the mixed solids remains
in the reactor. These solids make up the filter cake, and can
be returned to the bulk liquid through backwashing or cross‑flow
aeration.
|
same as
intermediate
|
Biological Activity
|
There is no
difference in the biological model when using simple, intermediate,
or advanced modes.
|
Model Parameters –
Physical
All physical and
operational parameters relating to the specification and operation
of the activated sludge reactors are identical to the CSTR and
plug-flow tank objects (with one exception – mechanical aeration is
not available). The specification of the membrane filter
physical characteristics is done in the Input Parameters >
Physical – Membrane menu, as shown in Figure
6‑15. All parameters except solids capture
rate are ignored (and greyed-out on these menus) when the model
is set to Simple Mode.
Figure 6‑15 – Physical – Membrane Forms
The GPS-X MBR model calculates the removal of solids due to the
membrane using a mass balance and the solids capture rate (actually
a fraction) specified in the
Input Parameters > Physical – Membrane form.
The default solids capture rate has been selected to provide
a permeate TSS concentration of 1 mg/L or less for typical MBR MLSS
concentrations (i.e. 8,000 to 12,000 mg/L). A separate mass
balance is applied to each particulate component in the biological
model.
In the simple mode, the permeate flow (and flux) is calculated
using a volumetric balance and the specified influent and pumped
flows. In advanced mode, the permeate flux through the
membrane is modelled using a resistance-in-series model (Choi
et al., 2000) as shown below:
Equation 6.53
where:
J
= permeate flux (m/s)
∆P =
trans-membrane pressure (kPa)
μ
= viscosity of water (Pa’s)
The trans-membrane pressure, ∆P, is specified in the
Input Parameters > Operational ‑Membrane
Form, as described in the section below. The
viscosity of water is calculated using the following equation
(Günder, 2001):
Equation 6.54
where:
T
= temperature (°C)
The intrinsic membrane resistance, Rm, is
provided in the
Input Parameters > Physical – Membrane
form. The default value used is based on data
provided by Chang et al. (1999). In advanced mode,
GPS-X tracks the formation of a cake layer on the surface of the
membrane that resists the liquid flux across the membrane.
The cake layer is assumed to be homogeneous and its thickness is
calculated as shown below in Equation 6.55 (Choi
et al., 2000).
Equation 6.55
where:
δc
= cake layer thickness (m)
mp
= dry mass of cake layer (kg)
pP
= density of a cake layer particle (kg/m3)
ε
= porosity of cake layer (dimensionless)
Am
= total membrane surface area (m2)
The dry mass of the cake layer (see Equation 6.54) is
calculated using the following dynamic mass balance on the cake
layer:
Equation 6.56
where:
qperm
= permeate flow rate (m3/d)
xliq
= concentration of solids in bulk liquid phase within the MBR
(kg/m3)
fcapture
= solids capture rate or fraction (dimensionless)
qbackw
= backwash flow rate (m3/d)
xcake
= mass of solids in cake layer (kg)
fbw
= backwash solids removal rate (1/m3)
qcross
= crossflow or air scour flow rate
(m3/d)
Am
= total membrane surface area (m2)
fcross
= cross-flow solids removal rate (kg/m)
Ks,cake
= half-saturation coefficient for cross-flow air
(kg)
The cake layer mass
balance considers the bulk convection of solids to the surface of
the membrane, the solids removed due to backwashing, and the solids
removed due to cross-flow aeration. Equation
6.55 ignores the diffusion away from the cake layer into
the bulk liquid as this term is assumed to be small compared to the
other terms in the mass balance. The half-saturation
coefficient for cross-flow air and the switching function based on
the mass of cake solids is used to smoothly stop the solids removal
as the cake layer disappears.
The default density
of a cake layer particle (see Equation 6.54) is equal
to the density of dry biofilm already used in the GPS-X fixed-film
reactors. The default porosity of the cake layer (see
Equation 6.54) is given in the
Input Parameters > Physical – Membrane > More…
form and is estimated using information given in Chang et
al. (1999). The membrane surface area
(see Equation 6.54) is
site-specific and should be selected to achieve a flux within the
range normally recommended by the manufacturer (e.g. a typical
range for hollow fiber membranes is 17 to 25 L/m2/h; see
Wallis-Lage et al., 2005).
The backwash
solids removal rate (see Equation 6.55) is
specified in the Input Parameters > Physical –
Membrane > More … form. It is
multiplied in Equation 6.55 by the backwash
flow and the cake solids mass
(term 2 in Equation 6.55) to give the
mass of solids removed from the filter cake per unit time.
The default backwash solids removal rate was calibrated
using data from Garcia and Kanj (2002). The
default backwash flow (see Equation 6.55) is
entered in the Input Parameters > Operational –
Membrane form and is based on data provided by Garcia
and Kanj (2002).
The cross-flow
solids removal rate (see Equation 6.55) is the
mass of solids removed from the filter cake, per unit of cross flow
air, per unit surface area of the filter. The default value
is entered in the Input Parameters > Operational –
Membrane > More … form and is calibrated based on data
in Garcia and Kanj (2002). The cross-flow
airflow (see Equation 6.55) is entered in the
Input Parameters > Operational –
Membrane form. The default value gives airflow per
surface area of 0.37 m3/m2/h
(see Wallis Lage et al., 2005)
for the default membrane surface area. The cross‑flow airflow
should be changed to reflect changes in the membrane surface
area.
The cake layer
resistance, Rc, is calculated by combining
Equation 6.54 and the Kozeny-Carman equation for flow
through porous passages as follows:
Equation 6.57
where:
dp
= effective cake
particle diameter (m)
The default effective cake
particle diameter is given in the Input
Parameters > Physical – Membrane >
More…form, and is estimated using information given
in Shin et al. (2002).
The fouling
resistance is calculated as follows (adapted from Choi et
al., 2000):
Equation 6.58
where:
Rf,max
= maximum fouling resistance (m-1)
kf
= fouling rate
constant (d-1)
t
= time since
last recovery clean (d)
The maximum
fouling resistance and fouling rate constant can be
found in the Input Parameters > Physical – Membrane
> form. They are selected based on data in
Merlo et al. (2000). The model assumes
that the fouling material is completely removed during a recovery
clean so that the time for fouling starts after each recovery
clean.
The membrane
operational parameters can be set in the
Input Parameters > Operational –
Membrane form, shown in Figure
6‑16.
Figure 6‑16 – Membrane Operational Parameters
Menu
The
trans-membrane pressure from Equation 6.1 is
set in this menu, along with the MBR backwashing options (including
the frequency, duration and flow rate of the backwash).
Alternatively, the level controller can be used, which sets
the backwash length to the amount of time required to refill the
tank. If the level controller is on and the tank is already
full, no backwashing will take place.
A helpful warning alarm can be
set in the Membrane Backwash …More… menu. The
“warn if tank is overflowing or empty” alarm will print
information to the Command Window if the filter flow causes the
final tank to overflow or become empty.
A membrane flux
controller can be activated to maintain the trans-membrane pressure
at the membrane flux setpoint using a built-in controller
with the rate of pressure increase being the controller
gain.
The cross-flow air
used to clean the membranes is assumed to be delivered using a
coarse-bubble aeration system. The alpha factor for
cross-flow air is based on total suspended solids
concentrations of between 8,000 mg/L and 10,000 mg/L in the
MBR. The standard oxygen transfer efficiency (cross-flow) is
based on a coarse bubble aeration system at submergence of 4.3 m
(14 ft.).
The cleaning
frequency sets how often physical/chemical cleaning is used to
reset the membrane fouling resistance to zero.
Output Variables
Several
membrane-specific output variables are available to be plotted. The
Membrane Filter Variables menu is accessed from the overflow
connection point (not the filter connection point), and is shown in
Figure 6‑17.
The MBR Cake
Variables menu (shown in Figure 6‑18) provides
output variables for total cake mass and cake
thickness.
Tips on Using the MBR
Models
The following
points are useful to keep in mind when setting up an MBR
simulation:
·
When modelling a recycle activated sludge (RAS) stream, the user
can simply use the internal recycle feature (located in
Operational – Tank > Internal Flow Distribution)
and set the internal recycle from the final (membrane) tank to the
desired tank.
·
Use Simple Mode initially to establish the required
operating conditions, such as MLSS, SRT, waste flow, etc.; and in
cases where details on the maintenance of the permeate flux are not
required. If you are only interested in the biological
treatment aspects of the system, use Simple Mode. If
you need to also simulate the physical aspects of the filter
operation (i.e. different backwash cycles, TMP management, etc.),
then Advanced Mode is required.
·
When using Advanced Mode, pay close attention to the volume
in the reactor(s), as it is easy to either drain or overfill the
reactor that contains the membrane filter.
·
If you wish to have a membrane “relaxation” period (rather than a
backwash period), set the duration and frequency as normal, but set
the backwash flow rate to zero. This will cause the permeate
flow to cease for the appropriate period of time, but there will be
no backwash flow into the tank. This can be done in both the
Intermediate and Advanced modes.
·
There is no steady-state solution for the model in Advanced
Mode. The discontinuous backwashing and cleaning cycles
render the model without a true steady equilibrium state (similar
to why the SBR object does not have a steady-state solution).
When using Advanced Mode, you may wish to run a long dynamic
simulation (~100 days) to allow time for the system to reach a
cyclic or periodic equilibrium. Due to the fact that there is
no backwashing or cleaning in Simple Mode, the steady-state
solver can be used.
A summary of the default values of the GPS-X MBR model parameters
are shown below in Table 6‑5:
Table 6‑5 –
GPS-X MBR Model – Default Parameter Values
Model Parameter
|
Unit
|
Default Value
|
Comment/Reference
|
Solids
capture
rate
|
-
|
0.9999
|
Default value
was selected to provide a permeate TSS concentration of 1 mg/L or
less for typical MBR MLSS concentrations
|
Density of
dry cake solids
|
kg/m3
|
1,020
|
Hydromantis
(2003)
|
Porosity of
cake layer
|
-
|
0.15
|
Chang et
al. (1999)
|
Solids
backwash removal rate
|
1/m3
|
100
|
Calibrated
using data in Garcia and Kanj (2002)
|
Cross-flow
solids removal rate
|
kg/m
|
200,000
|
Calibrated
using data in Garcia and Kanj (2002)
|
Intrinsic
membrane resistance
|
1/m
|
1.0e+11
|
Chang et
al. (1999)
|
Maximum
fouling resistance
|
-
|
1.0e+12
|
Calibrated
using data in Merlo et al. (2000)
|
Fouling rate
constant
|
1/d
|
0.005
|
Calibrated
using data in Merlo et al. (2000)
|
Figure 6‑17 - Membrane Output Variables
Menu
Figure 6‑18 - MBR Cake Variables Menu
Model Structure
The anaerobic MBR
model in GPS-X combines the completely-mixed MBR (in simple mode)
with the gas transfer and headspace model of the anaerobic digester
object. The result is a completely mixed membrane bioreactor
with a closed headspace and gas production. Unlike the
completely-mixed MBR, there is no modelling of the cake formation
on the membrane surface, backwashing of solids, or trans-membrane
pressure calculations.
The model assumes
that the reactor is completely mixed with no aeration, and that
filtrate flow from the MBR is equal to the influent flow minus the
pumped flow. It is assumed that the filter is not limiting to
the flow (i.e. that whatever flow is calculated can pass through
the filter unimpeded). The solids captured by the filter are
retained in the completely mixed bulk liquid.
The anaerobic MBR
is only available in the Selenium and Sulphur (Mantis2) and the
Carbon Footprint (Mantis3) libraries. The anaerobic
biological and chemical reactions are identical to those used in
the anaerobic digester and UASB models.
Model Input
Parameters
The physical menu
of the anaerobic MBR object is used to specify the dimensions of
the liquid tank and the headspace. In addition, the headspace
total gas pressure, gas-liquid transfer constant, and
membrane solids capture are specified here as well, as shown
in Figure 6‑19 below.
Figure 6‑19
– Anaerobic MBR Physical Parameters Menu
In the Operational
input parameters menu, the only operational parameter to be
considered is the pump flow rate from the pumped connection on the
lower right-hand corner of the object. A PID control loop is
available to be configured as well.
The remainder of
the menus are similar to those shown in other completely mixed
biological reactors.
Model Output
Variables
The output
variables available for the anaerobic MBR object are a combination
of those from the conventional CSTR and the anaerobic
digester. The filtrate and pump connection points show flows,
concentrations and operating cost variables for the liquid
streams. The gas connection (at the top of the object) shows
the gas flow and composition.
Common Features
The models
associated with the sequencing batch reactor (SBR) object are
combinations of suspended-growth and sedimentation models. The
various aerated and mixed phases use a suspended-growth model,
assuming a completely mixed hydraulic configuration, while the
settling and decanting phases use a reactive sedimentation model.
The models are combined together to form the whole unit process
model.
Modes of Operation
There are three different SBR objects in GPS-X: the simple
sequencing batch reactor (SBR) object, the Advanced SBR object, and
the Manual SBR object. All three objects have the same
functionality, appearance and choice of biological models.
They differ in the manner in which the user specifies the operation
of the SBR unit.
The simple and advanced SBR objects require the
specification of the timing and flow rates to define the phases.
The manual SBR object requires that the entire operational
cycle be defined by the user, either by having the liquid flows,
air flows, and mixing on interactive controllers, or as file
inputs. The manual object is suited to operator training,
while the simple and advanced objects are typical for
an SBR application.
When using multiple, parallel
SBRs, it may be desirable to stagger the cycle times of the units.
For example, if two parallel SBRs are in use, each with a 6‑hour
cycle time, it may be desired to shift one of the SBRs by 3 hours.
If both have 3 hours of detention time, then a continuous
influent could be switched between the two unit processes. To
specify a timeshift for an SBR model using either the simple or
advanced objects, the user enters the desired timeshift for the
timeshift for simple and advanced cycles parameter. The
timeshift parameter cannot be set to a value greater than the cycle
time.
Simple SBR Cycle
The parameters used
to specify the operational cycle (in the regular SBR object, not
advanced or manual) are found in the Parameters >
Operational – Cycle Settings menu. These parameters
include the duration of one complete cycle and associated seven
separate phases, which have predefined functions. The parameters
are shown in Figure 6‑20.
The seven fixed
phases for each cycle are described below Figure
6‑20.
Figure 6‑20 - Regular SBR - Operation Cycle
Parameters
1.
Mix (and fill): The unit is modelled as a CSTR without
aeration (i.e. air flow, power or KLa set to 0.0). This
phase is usually the starting phase, used while the tank has just
started filling with liquid. This phase represents the first of
four mixing phases.
2.
Aerate (and fill): The unit is modelled as a CSTR with
aeration. This represents the second mixing phase, and is used
while the tank is filling after the aeration has been turned
on.
3.
Mix only: The unit is modelled as a CSTR without aeration.
This represents the third mixing phase usually occurring sometime
after the tank has finished filling. It is typically used as a
denitrifying phase.
4.
Aerate:The unit is modelled as a CSTR with aeration. This is
the last mixing phase generally used to re-aerate the sludge so
that it will settle properly.
5.
Settle:The unit is modelled as a settler, with or without
biological reactions, depending on the model selected. The mixing
is turned off to allow the content of the tank to become quiescent
and to promote settling.
6.
Decant: The unit is modelled as a settler. During this
phase, the user-specified decant flow is activated.
7.
Desludge:The unit is modelled as a settler. This phase
occurs at the end of the cycle, and the user-specified wastage flow
is activated.
With the normal SBR object, the seven phases making up this cycle
are fixed in sequence (the order is shown in the form). The total
length of all seven phases must not exceed the cycle time
specified. If the total length of all seven phases adds up to less
than the cycle time, the tank will be idle until the end of the
cycle time. If it adds up to a value greater than the cycle time,
the remaining time and/or phases beyond the cycle time will be
ignored. A phase can be disabled by setting its length to
zero.
Figure 6‑21 - Advanced SBR Operational
Parameter
The decant flow rate (pumped flow parameter) and waste flow
rate (underflow rate parameter) are entered in the
Parameters > Operational - Flow Control form.
The influent flow to the SBR is
taken from the upstream object (e.g. influent object) if the
influent #1 pump label parameter is left blank in the
Flow Control form. The user must then make sure that the
influent flow is synchronized with the SBR cycle (i.e. entering the
SBR during the appropriate phase). Alternatively, in the Flow
Control form, you can enter a stream label for the influent
#1 pump label parameter, a flow rate for the first influent
flow parameter, and the SBR influent flow will be automatically
taken from that stream during the appropriate phase within the
cycle (automatic synchronization). The stream label must be
associated with a pumped flow stream (`qcon' variable), for
example an influent icon or pumped flow from a tank. The
influent #2 pump label parameter is ignored in this cycle
type.
Advanced SBR Cycle
Settings
The SBR operational
cycle parameters for the Advanced SBR model are shown in
Figure 6‑21. The advanced SBR cycle is
more general than the simple SBR model, allowing the user to define
up to ten or more different phases. For each phase defined, the
corresponding duration, mixing (either on or off), aeration (oxygen
mass transfer coefficient), decant flow rate, and wastage flow rate
can be specified. The order of the user-defined phases is fixed
according to their order presented in the forms.
The influent flow to the SBR is
taken from the upstream object (e.g. influent object) if the
influent #1 pump label and influent #2 pump
labelparameters are left blank in the Flow Control form.
The user must then make sure that the influent flow is synchronized
with the SBR cycle (i.e. entering the SBR during the appropriate
phase). Alternatively, you can enter stream labels for the
influent #1 pump label and influent #2 pump label
parameters in the Flow Control form, flow rates for the
influent #1 flow in phase and influent #2 in phase
parameters in the form (Advanced section), and the SBR
influent flow will be synchronized automatically.
The stream labels must be
associated with pumped flow streams (`qcon...' variables),
for example an influent icon or pumped flow from a tank. Most
applications will only use one influent, but second influent was
provided for cases were methanol is added, for example.
Manual SBR Model
If the manual SBR
object is used, the user must set up all the important parameters
such as air flow rate, mixing, decant flow rate, and waste flow
rate on interactive controllers to change them as the simulation
proceeds.
Alternatively, the
parameters can be set up as file inputs or controlled from custom
code in the .usr
file.
Figure 6‑22
- Manual Cycle Operational Parameters
In the manual SBR model, there are no cycle settings as in the
simple and advanced models. The timing of mixing, aeration
and pumping all must be controlled directly.
The decant flow rate (pumped flow parameter) and waste flow
rate (underflow rate parameter) can be entered in the
Parameters > Operational - Flow Control form.
The influent flow to the
SBR is taken from the upstream object (e.g. influent object) if the
influent #1 pump label parameter is left blank in the
Flow Control form. The user must then make sure that the
influent flow is synchronized with the SBR cycle
(i.e. entering the SBR during the appropriate phase).
Alternatively, in the Flow Control form, you can enter a
stream label for the influent #1 pump label parameter, a
flow rate for the first influent flow parameter, and the SBR
influent flow will be automatically taken from that stream during
the appropriate phase within the cycle (automatic
synchronization). The stream label must be associated with a
pumped flow stream (`qcon' variable), for example an
influent icon or pumped flow from a tank. The influent #2 pump
label parameter is ignored in this cycle type.
The pond object
contains only the empiric model, which simulates the
transformation, dilution and mixing of state variables in a pond.
Unlike the activated sludge models, the empiric model is
empirical in nature. It does not use fundamental mechanistic
dynamic processes to determine the rate of change of the states
(such as those found in the Model Matrices in Appendix A), but
simulates behavior that has been observed in existing ponds,
bench‑scale studies and pilot-scale studies.
The empiric model simulates
three different kinds of ponds: anaerobic,
facultative, and aerated. The pond models are
discussed in detail below.
Note that
Pond/Lagoon object is only available in the GPS-X legacy libraries
(cnlib, cniplib, cnplib, cnpiplib)
Anaerobic Ponds
In the anaerobic pond model, the BOD and TSS removal is simulated
using a simple regression model derived from anaerobic pond
behavior in North America (Beier, 1987).
The following equations describe
the amount of BOD and TSS removal:
Equation 6.59
Equation 6.60
where:
ATSS, ABOD, BTSS,
BBOD, Khrt = calibration parameters
n =
number of ponds in series
As the empirical equations above only deal with GPS-X composite
variables (BOD and TSS), a methodology is required to transfer this
information back to the state variables. The following list
outlines the relationships for the state variables in CNLIB.
A fraction is calculated for each particulate state variable
(xbh, xba, xs, xi, xu, xsto for CNLIB) that is equal to that
variable's concentration over incoming xcod.
A new TSS value (x) is
calculated from the above empirical REMOVAL equation.
New VSS (vss) and particulate COD (xcod)
concentrations are calculated as:
vss = ivt
* x
xcod = icv * vss
A new BOD value is
calculated from the above empirical REMOVAL equation, and
then compared to the BOD value calculated as the sum of the
fractions from step 1 multiplied by the new xcod multiplied
by fbod.
If the BOD calculated from the fractions is less than the BOD
calculated from the empirical equation, excess soluble BOD (the
difference between bod and xbod) is assigned to
ss.
If the BOD calculated in
step 3 from the fractions is greater than the BOD calculated from
the empirical equation, the excess particulate BOD (xbodu)
is assumed to be converted to xu. Soluble substrate
(and soluble BOD) is then assumed to be zero.
Inert and non-reacting variables are mapped through the pond object
(si, salk, and snn).
The effluent oxygen
concentration (so) and nitrate concentration (sno)
are assumed to be zero in the anaerobic pond environment.
Particulate biodegradable organic nitrogen (xnd) is assumed
to be converted to snh as per the conversion of
xs.
Soluble biodegradable
organic nitrogen (snd) is assumed to be converted to ammonia
(snh) as per the conversion of ss.
The loss of heterotrophic
(xbh) and autotrophic (xba) biomass results in the
production of ammonia (snh).
Facultative Ponds
The approach for
facultative ponds is similar to that for anaerobic ponds, but uses
a different model for BOD reduction, and a slightly different
approach for determining state variable concentrations. The
empirical BOD removal model used for facultative ponds is from
Thirumurthi (1974):
Equation 6.61
where:
BODEFF
= effluent BOD (g/m3)
BODINF
= influent BOD (g/m3)
KS
= first order BOD removal rate coefficient (1/d) at 20°C
CTEMP
= correction factor for temperature
CO
= correction factor for organic loading
and
Equation 6.62
Equation 6.63
where:
TEMP
= temperature, °C
θ
= temperature correction constant
SLR
= standard loading rate (kg/ha/d)
LR
= current loading rate (kgBOD/ha/d)
The methodology for the
determination of the state variables xba, xbh, xs, xu, xi, xsto,
si, salk, snn, ss, snd and xnd are identical to that for
anaerobic ponds.
Oxygen is determined by assuming
that the DO is saturated in the aerobic zone, and zero in the
anaerobic zone. The concentration of so
is determined from the following equation:
Equation 6.64
where:
SOST
= saturated oxygen concentration (gO2m3)
AERDEPTH = fraction of depth that is
aerobic (unitless)
For the nitrogen
composite variables, nitrate is assumed to be converted to nitrogen
gas in the anaerobic zone; therefore:
Equation 6.65
As anaerobic
conversion of biomass and biodegradable nitrogen only happens in a
fraction of the pond, the increase in ammonia (snh) as shown
in the anaerobic model is multiplied by AERDEPTH.
Aerated Ponds
The aerated pond uses the same empirical BOD and TSS reduction
models as the facultative pond model. However, the default value
for the BOD removal rate coefficient has been changed to reflect
aerated conditions (Eckenfelder, 1980).
The methodology for the determination of the state variables
xba, xbh, xs, xu, xi, xsto, si, salk, snn, ss, snd and
xnd is identical to that for facultative ponds.
Oxygen is assumed to be
saturated at the effluent point of the aerated pond. Due to this
completely oxic environment, nitrate is assumed to remain
unchanged, and is mapped through the pond object.
Ammonia is assumed not to
change, due to the assumption that most of the biomass will be
settled out of the water column. Therefore, snh is mapped
through the pond object.
Calibration
The parameters
found in the Parameters > Physical page of the pond
object include pond type and settings for pond size.
Number of cells in series describes the degree of plug flow
for the pond system. Typically, BOD and TSS removal increases with
increasing numbers of cells in series. The temperature values found
on the More... page are used in the temperature correction
equations in the facultative and aerated pond models.
The parameters
found on the Parameters > Empirical Model
Constants include those used to calibrate the BOD and TSS
removal, and the half-saturation coefficient KHRT
that is used to decrease removal efficiency for very short
HRTs.
Table 6‑6 summarizes
suggested calibration techniques for the empirical pond models
found in the empiric model.
Table 6‑6 – Calibration Suggestions
Model
|
Calibration
|
anaerobic
|
Calibrate with the A coefficients
for TSS removal first, then adjust A coefficient for BOD.
|
facultative
|
Calibrate with the A coefficients
for TSS removal first, then adjust BOD removal rate constant.
Further tuning can be performed by adjusting KHRT and temp/loading
coefficients. Lastly, adjust depth of aerobic zone to calibrate
ammonia.
|
aerated
|
Calibrate with the A coefficients
for TSS removal first, then adjust BOD removal rate constant.
Further tuning can be performed by adjusting KHRT and temp/loading
coefficients.
|
Pond Models in Other Libraries
The above methods
for determining the values of state variables from the empirical
pond models are for CNLIB only. These methods have been extended to
the other libraries by making small additions and changes based on
the new and/or different state variables. Table 6‑7
describes these differences.
Table 6‑7 -
Library-specific Algorithms for Empiric Pond Model
Library
|
Additions/Changes from CNLIB pond
methodology
|
CNIPLIB
|
All particulate IP components (xza to
xzo) settle
similarly to xi (not changed biologically, but undergo settling and
mixing, as appropriate). Soluble IP components (sza to
szo) are
unaffected by biological activity, but undergo mixing.
|
CNPLIB
|
slf and
sf behave
similarly to ss. PAOs are
transformed similarly to xbh
and xba.
sp is not
modelled, and is treated the same as si.
|
CNPIPLIB
|
slf and
sf behave
similarly to ss. PAOs are
transformed similarly to xbh
and xba.
sp is not
modelled, and is treated the same as si. All
particulate IP components (xza to
xzo) settle
similarly to xi (not changed
biologically, but undergo settling and mixing, as appropriate).
Soluble IP components (sza to
szo) are
unaffected by biological activity, but undergo mixing.
|
Setting the Recycle
Flow
The oxidation ditch
object operation as 16 CSTRs in series (a plug flow tank) with a
large recycle from the last tank to the first. The recycle
can be specified in four different ways from the Operational
menu, using the ditch recirculation mode setting, shown in
Figure 6‑23.
Figure 6‑23 - Oxidation Ditch Recirculation
Mode Settings
·
Set ditch velocity – specify a constant surface velocity of
the ditch flow
·
Set constant ditch flowrate – rather than specifying
velocity, set the flow
·
Set proportional ditch flowrate – make the ditch flow
proportional to the incoming flow to the oxidation ditch.
Typically, the ditch flow would be 50 to 200 times the incoming
flow.
·
Set constant outlet fraction – specify the fraction of the
ditch flow that exits the ditch each time around.
Special 2-D Greyscale
Output Variables
The oxidation ditch
has a series of special two-dimensional output variables that are
desiged to be displayed on 2-D greyscale outputs. These
variables are found in the 2-D Greyscale Ditch Output output
variables menu. These graphs show a plan-view of the
oxidation ditch, with tank #1 in the lower right-hand corner, and
the effluent point in the upper-right hand corner (the actual
influent point depends on the setting in the influent fractions
menu). The flow moves clockwise around each ditch.
Five different
variables are available for output: Dissolved oxygen (DO),
Ammonia, Nitrate, Oxygen uptake rate (OUR) and, Denitrification
rate (DNR)
Figure 6‑24 shows an example of the 2-D greyscale DO
graph for a typical oxidation ditch.
Figure 6‑24 - 2-D Greyscale Oxidation Ditch
Output - Dissolved Oxygen
Please Note: Due to the high recycle rate
within the object (and subsequent short HRTs for each section of
the ditch), the oxidation ditch model often will take longer than
normal to converge to steady-state with the steady-state
solver. You may find it necessary to increase the
iteration termination criteria and/or the damping factor
on final approach in the steady-state solver menu to achieve a
reasonable solution.
If you are having
difficulty with steady-state solutions, you can halt the
steady-state solver at a higher iteration termination value, and
run a short dynamic simulation to test if the steady-state solver
solution is adequate. If the dynamic solution does not
diverge appreciably from the steady-state conditions, then the
higher iteration termination value is suitable. If you have
any questions or problems, please contact us for assistance at
support@hydromantis.com
Common Features
The models
associated with the continuous flow sequencing reactor (CFSR)
object operates as set of CSTRs in series (a plug flow tank) with a
large recycle from the last tank in the series to the first tank.
The recirculation flow rate can be specified in 5 different ways
from the operational menu, using the recirculation
mode setting, shown in Figure 6‑25. The
flow moves clockwise around the reactor, in ascending order of CSTR
number.
Figure 6‑25 - Continuous Flow Sequencing
Reactor Recirculation Settings
·
Set recirculation velocity – specify a constant surface
velocity of CFSR flow
·
Set constant recirculation flowrate – rather thatn
specifying velocity, set the flow
·
Set proportional recirculation flowrate – make the CFSR flow
proportional to the incoming flow to the CFSR.
·
Set constant outlet fraction – specify the fraction of the
flow that exits the CFSR each time around.
·
Set proportional to bridge – set the recirculation rate
proportional to the rotational period of the aeration bridge
(simulate rotaing aeration brigde does not have to be on for this
option to function)
Please
Note: Due to the high recycle rate
within the object (and subsequent short HRTs for each section of
the reactor), the CSR model often will take longer or fail to
converge to steady-state with the steady-state solver. You may find
it necessary to increase the iteration termination criteria in the
steady-state solver menu to achieve a reasonable
solution.
If you are having
difficulty with steady-state solutions, you can halt the
steady-state solver at a higher iteration termination value, and
run a short dynamic simulation to test if the steady-state solver
solution is adequate. If the dynamic solution does not diverge
appreciably from the steady-state conditions, then the higher
iteration termination value is suitable. If you have any questions
or problems, please contact us for assistance at
support@hydromantis.com.
Rotating Aeration
Bridge
The CFSR consists
of a circular tank with an aeration grid suspended from a rotating
bridge, though stationary aeration grids can be installed on the
floor along the perimeter of the tank. When modelling the rotating
aeration bridge, airflow specified in the total airflow to rotating
diffusers is moved around the tank from one CSTR cell to the next
over the specified rotational period. The rotational period of
aeration bridge specifies the period of time required for the
rotating diffusers to make one revolution around the tank.
Figure 6‑26 - Continuous Flow Sequencing
Reactor Rotating Aeration Bridge Setting
Controller Operation
The CFSR object
specifies oxygen transfer by using either a DO controller option or
entering airflow as the manipulated variable in place of
KLa (similar to other DO controllers in GPS‑X).
The PID DO controller manipulates KLa and back
calculates the required airflow.
The CFSR object
contains one additional specify oxygen transfer by...
option: The DO On/Off Controller is switches the airflow
into the tank on and off depending on the specified DO high and low
limits. How quickly action is taken by the controller (after the
upper or lower limit is reached) is dependent on the controller’s
specified sampling time.
The CFSR object in
GPS-X contains three additional Aeration Controllers which
can be paired with any of the specify oxygen transfer by...
options: timer, nitrate controller and the ammonia
controller. These additional controllers can be seen in
Figure 6‑27. While the specify
oxygen transfer by... controller options control only the DO
concentration, the aeration controllers manage more complex
aeration strategies in the tank. During aerated phases of
each of the aeration controllers, the selected specify oxygen
transfer by... option manipulates the airflow according to the
following strategies:
Figure 6‑27 - Continuously Flow Sequencing
Reactor Additional Aeration Controllers
·
Timer – The timer controller creates a fixed cycle of timed
aerated and un-aeratred phases. The timer controller is
configured by specifying the aeration start time, the aeration end
time, and the length of the entire cycle, where the remaining time
in the cycle is un-aerated.
·
Ammonia controller – The ammonia controller controls airflow
by turning on the air when the ammonia reaches the ammonia high
limit in anoxic phase, and turns off the air when the ammonia
reaches the ammonia low limit in oxic phase.
·
Nitrate controller – The nitrate controller consists of
three phases: Oxic, anoxic, and anaerobic:
o
Oxic – During the oxic phase, the aeration system in the
CFSR is on. Aeration in the tank continues until the
nitrate high limit in oxic phase, nitrate concentration, is
reached. Upon reaching the specified upper concentration
limit, the controller switches to the anoxic phase.
o
Anoxic – During the anoxic phase, the aeration system in the
CFSR is off. The anoxic phase continues until the nitrate
low limit in anoxic phase is reached. Upon reaching the
specified low concentration, the controller switches to the
anaerobic phase. If the average nitrate removal rate across
the CSR drops below the minimum nitrate removal rate in anoxic
phase before reaching the low nitrate concentration, the
controller will end the anoxic phase and return directly to the
oxic phase without executing the anaerobic phase.
o
Anaerobic – The anaerobic phase is a time phase of specified
duration, commencing at the completion of the anoxic phase.
During the anaerobic phase, the aeration system in the CFSR is
off. Once the length of time specified in the anaerobic
phase length has passed, the controller will return to the oxic
phase.
The high purity
oxygen (HPO) activated sludge object works the same way as a
plug-flow tank object, except that the aeration system uses a HPO
gas feed instead of regular air. In addition, the entire plug
flow system is capped, and each reactor has a headspace. The
headspaces are connected, to allow downstream flow. The flow
of gas downstream is determined from the gas input and venting
flows, and equalizes the pressure in all the reactor
headspaces. Figure 6‑28 shows the physical
configuration of the HPO system
Figure 6‑28 - Schematic of High Purity Oxygen
(HPO) System
The flow of gas and flow of water are modelled separately.
The exchange of O2 gas, N2 gas and
CO2 gas at the air/water interface is determined using
Henry’s Law, corrected for temperature and headspace pressure.
The regular biological models have been supplemented in the HPO
object with stoichiometry to calculate CO2 gas
generation. As the gas travels downstream, O2 is
transferred into the liquid and consumed by the biological
activity. N2 and CO2 are generated, and
equilibrated with the gas concentrations in each headspace.
Consequently, the composition of the gas (fraction of the gas that
is O2, N2 and CO2) can be
displayed for each reactor in the plug-flow system. Output
variables for the gas composition can be found in the HPO System
Variables output variables menu.
As the HPO system is capped, and CO2 gas is contained in
the headspace at concentrations often much higher than atmospheric
values, HPO systems may have atypical pH. The GPS-X HPO
object contains a pH calculator that determines the pH in the
liquid for each reactor. For details on the pH model, refer
to the pH discussion in the “Tools” chapter (the model in the
Toolbox is the same one used here). It is important to
properly specify the anion and cation concentrations in the tank to
achieve a calibrated pH calculation.
The pH that is determined for each reactor can be set to inhibit
biological growth, using the pH inhibition settings in the
Physicalparameters more… button. The pH
inhibition is set to OFF by default.
The growth of biomass is multiplied by a pH inhibition factor,
taken from
Grady and Lim (1980), which is
bounded between zero and one. This creates a linear decline
between pH = 7.2 and pH ≈ 6.1.
Equation 6.66
Open Basic HPO (MANTIS2LIB
and MANTIX3LIB library only)
The open basin HPO
unit process models the gas-liquid transfer processes for a pure
oxygen fed activated sludge process. The open basin systems are
considered to provide good oxygen transfer efficiency while
allowing better exchange of CO2. The better exchange of CO2
prevents excessive drop in pH. The open basin HPO model has the
following features:
1.
A gas-liquid transfer model based on feed gas and outlet gas
composition
2.
Temperature estimation model based on energy balance
3.
High temperature and kinetic parameter relationship
Typical outputs
from the energy balance model and the oxygen transfer model are
shown in Figure 6‑29 and Figure
6‑30.
Figure 6‑29 - Typical Outputs from the Energy
Balance Model for Temperature Estimation
Figure 6‑30 - Typical Outputs for the Oxygen
Transfer Rate
Modelling of Gas Transfer
in Open Basin HPO
The gas transfer to
the bulk liquid phase of a biological reactor is modelled using a
dynamic mass balance written for each dissolved gas
(Hydromantis, 2011). For example, a dissolved
oxygen mass balance around a completely stirred tank reactor (CSTR)
is shown below.
Equation 6.67
where:
V = reactor
volume (m3)
CL = concentration of
dissolved oxygen (DO) in the reactor (mg/L)
Q =
influent flow rate (m3/d)
Cin =
concentration of DO entering reactor (mg/L)
KLa = oxygen mass
transfer coefficient at field conditions (1/day)
C*∞ =
DO saturation concentration at field conditions (mg/L)
r
= rate of use of DO by biomass (g/day), the respiration rate
The volume, flows,
and reaction rates are known from specifications or other modelling
equations, leaving two terms that must be calculated in order to
solve the dissolved oxygen mass balance over time for the DO
concentration in the reactor, CL:
1.
DO saturation concentration at field conditions,
C*∞, and
2.
Oxygen mass transfer coefficient at field
conditions, KLa
Calculation of DO
Saturation Concentration at Field Conditions
The DO saturation
concentration at field conditions is calculated as follows:
Equation 6.68
where:
τ
= temperature correction factor (unitless)
β
= correction factor for sales, particulates, and surface-active
substances (unitless)
Ω
= pressure correction factor (unitless)
C*∞20
= DO saturation concentration at 20°C and 1
atm (mg/L)
The correction factors are used to adjust the DO
saturation concentration to account for the temperature of the
liquid, the pressure at the submergence level of the diffusers, and
the salts, precipitates, and surface-active substances found in the
wastewater.
The temperature correction factor is calculated as
follows:
Equation 6.69
where:
C*st
= surface DO saturation concentration at temperature of t and 1 atm
air pressure (mg/L)
C*s20
= surface DO saturation concentration at 20°C and 1 atm pressure
(mg/L)
The surface DO saturation concentration at liquid temperature t and
an air pressure of 1 atm is obtained using a look-up table in
GPS-X that is based on temperature. The look-up table data
were taken from Appendix C of the USEPA Design Manual – Fine
Pore Aeration Systems (USEPA, 1989). When
the temperature falls between two data points in the table, GPS-X
uses linear interpolation to determine the
C*st value. The value
of C*s20 is 9.09 mg/L.
The correction
factor for salts, particulates, and surface-active substances,
β, is a parameter that must be measured or estimated for the
wastewater of interest. In GPS-X, a default value of 0.95 is
used.
The pressure correction factor is calculated as follows:
Equation 6.70
where:
Pb
= barometric pressure at elevation and air temperature (kPa)
Ps
= standard barometric pressure (101.325 kPa)
pv
= vapour pressure of water at liquid temperature (kPa)
The barometric pressure at elevation and air temperature is
calculated using the following formula taken from Appendix B-2 of
Metcalf and Eddy (2003):
Equation 6.71
where:
g =
acceleration due to gravity (9.81 m/s2)
M = molecular
weight of air (28.964 kg/kg-mole)
R =
universal gas constant (8314 N•m/kg-mole•K)
Tair = air
temperature (K)
zi
= elevation at
position “i” (m)
The vapour pressure of water at the liquid temperature is
determined using the Antoine equation (Felder and Rousseau,
1986):
Equation 6.72
where:
A, B, C
= Antoine coefficients
T
= wastewater temperature (°C)
The depth correction factor for oxygen saturation is calculated as
follows:
Equation 6.73
where:
d =
depth of the tank
The DO saturation concentration at 20°C and 1 atm is calculated
as follows:
Equation 6.74
The η for each gas is estimated using the following expression:
Equation 6.75
where:
molfro2
= average mole fraction of oxygen in bubble (-)
molfratmo2 = mole
fraction of oxygen in air (-)
The average mole
fraction of oxygen in the bubble is estimated as below:
Equation 6.76
where:
molfrffeedO2
= mole fraction of oxygen in feed gas (-)
molfrfoutO2
= mole fraction of
oxygen in out gas (-)
wffeed
= weight factor for averaging (-, default values is 1.0)
The final expression for the DO saturation concentration at field
conditions is as follows:
Equation 6.77
Oxygen Mass Transfer
Coefficient at Field Conditions Calculations
GPS-X provides four
different ways to specify the user inputs to estimate gas-liquid
transfer in the open basin HPO system.
1.
Direct specification of the clean water at KLa at
20°C
2.
Specifying the gas flow rate at standard condition of 1 atm
pressure and 20°C temperature
3.
Specify the wire point input of the oxygenator
4.
Use a DO controller
In each of the
above, the user has the option of specifying SOTE. The
default value for SOTE for high purity oxygen is considered to be
90%.
Entering KLa
Directly
Estimate
KLa at field conditions
The KLa at field
conditions is calculated using Equation 6.11:
Equation 6.78
where:
KLaT
= mass transfer coefficient at temperature T in °C (1/day)
KLa20
= mass transfer coefficient at 20°C
θ
= temperature correction factor (default value
GPS‑X is 1.024)
α
= wastewater correction factor
for KLa20
F
=diffuser fouling factor
(default value in GPS-X is 1.0)
T
= wastewater temperature (°C)
Estimate Oxygen
Transfer Rate (OTR) at Field Conditions in g/d
Equation 6.79
Estimate
Standard Oxygen Transfer Rate (SOTR) in g/d
Equation 6.80
Airflow at
Standard Conditions in m3/d
The gas flow is
calculated usingEquation 6.81:
Equation 6.81
where:
molfrO2
= mole fraction of O2 in user-defined air
(mole/mole)
MWO2
= molecular weight of O2 (32 g/mole)
Puser
= density of user-defined air (g/m3)
MWuser
= average molecular weight of user-defined air (g/mole)
Estimate Airflow
rate at Field Conditions
The airflow at standard conditions is converted to field conditions
using the ideal gas law:
Equation 6.82
where:
Tstandard
= air temperature at standard conditions (°C)
Tfield
= air temperature at field conditions (°C)
Estimate the
SAE
The SAE is estimated based on the correlation provided by
Praxair. The expression is valid for SOTE in the range
of 86% to 100%. If the user specified SOTE is less than 80%,
then SAE is bounded at 7.11 kg/kWh.
Equation 6.83
where:
SAE = Specific aerator energy
(wire), kgO2/kW-hr
SOTE = Standard oxygen transfer rate (%)
The mechanical power requirement is estimated using the following
expression:
Equation 6.84
where:
Pmechanical
= mechanical power (kW)
SAE
= mechanical aerator oxygen transfer rate (kg O2/kW·h)
CF2
= conversion factor (24,000)
ηmotor
= motor efficiency
Entering Airflow
The gas flow rate
is entered at a given standard condition of 1 atm and 20°C.
Standard Oxygen
Transfer Rate (SOTE) in g/d
The SOTR is calculated using Equation 6.85:
Equation 6.85
where:
molfrO2
= mole fraction of O2 in feed gas (mole
O2/mole gas)
MWO2
= molecular weight of O2 (32 g/mole O2)
Pgas
= density of feed gas (g/m3)
MWgas
= average molecular weight of gas (g/mole gas)
Estimate
OTR
The OTR is calculated using Equation 6.86:
Equation 6.86
Estimate
KLa at Standard Condition
Equation 6.87
Estimate
KLa at Field Condition
Equation 6.88
Airflow at
Standard Conditions in m3/d
The gas flow is
calculated usingthe following expression:
Equation 6.89
where:
molfro2
= mole fraction of O2 in feed gas (mole
O2/mole gas)
MWO2
= molecular weight of O2 (32 g/mole O2)
Pgas
= density of feed gas (g/m3)
MWgas
= average molecular weight of gas (g/mole gas)
Estimate Airflow
Rate at Field Conditions
The airflow at standard conditions is converted to field conditions
using the ideal gas law:
Equation 6.90
where:
Tstandard
= air temperature at standard conditions
Tfield
= air temperature at field conditions (°C)
Estimate the
SAE
The SAE is estimated based on the correlation provided by Praxair.
The expression is valid for SOTE in the range of 86% to 100%. If
the user specified SOTE is less than 80% then the SAE is bounded at
7.11 kg/kWh.
Equation 6.91
where:
SAE = Specific aerator energy
(wire), kgO2/kW-hr
SOTE = Standard oxygen transfer rate (%)
The mechanical power requirement is estimated using the following
expression:
Equation 6.92
where:
Pmechanical
= mechanical power (kW)
SAE
= mechanical aerator oxygen transfer rate (kg O2/kW·h)
CF2
= conversion factor (24,000)
ηmotor
= motor efficiency
Entering Oxygenator
Power
Estimate the
SAE
The SAE is
estimated based on the correlation provided by Praxair. The
expression is valid for SOTE in the range of 86% to 100%. If the
user specified SOTE is less than 80% then the SAE is bounded at
7.11 kg/kWh.
Equation 6.93
where:
SAE = Specific aerator energy
(wire), kgO2/kW-hr
SOTE = Standard oxygen transfer rate (%)
Estimate the
SOTR
The mechanical power requirement is estimated using the following
expression:
Equation 6.94
where:
Pmechanical
= mechanical power (kW)
SAE
= mechanical aerator oxygen transfer rate (kg O2/kW
h)
CF2
= conversion factor
ηmotor
= motor efficiency
Estimate
OTR
The OTR is
calculated using Equation 6.95:
Equation 6.95
Estimate
KLa at Standard Condition
Equation 6.96
Estimate
KLa at Field Condition
Equation 6.97
Airflow at Standard
Conditions in m3d
The gas flow is
calculated using Equation 6.98:
Equation 6.98
where:
molfro2
= mole fraction of O2 in feed gas (mole
O2/mole gas)
MWO2
= molecular weight of O2 (32 g/mole O2)
Pgas
= density of feed gas (g/m3)
MWgas
= average molecular weight of gas (g/mole gas)
Estimate Airflow
rate at Field Conditions
The airflow at
standard condition is converted to field conditions using the ideal
gas law:
Equation 6.99
where:
Tstandard
= air temperature at standard conditions
Tfield
= air temperature at field conditions (°C)
DO Control
Estimate
KLa at field condition
The
KLa at field conditions is calculated by the
controller
Estimate
KLa at standard condition
Equation 6.100
where:
KLaT
= mass transfer coefficient at temperature T in °C (1/day)
KLa20
= mass transfer coefficient at 20°C
q
(T-20)
= temperature correction factor (default value in GPS-X is
1.024)
aF
= wastewater
correction factor for
KLa20
F
= fouling factor (default value in GPS-X is 1.0)
T
= wastewater temperature
Estimate Oxygen
Transfer Rate (OTR) at Field Conditions in g/d
Equation 6.101
Estimate
Standard Oxygen Transfer Rate (SOTR) in g/d
Equation 6.102
Airflow at
Standard Conditions in m3d
The gas flow is calculated using Equation 6.103:
Equation 6.103
where:
molfro2
= mole fraction of O2 in feed gas (mole
O2/mole gas)
MWO2
= molecular weight of O2 (32 g/mole O2)
Pgas
= density of feed gas (g/m3)
MWgas
= average molecular weight of gas (g/mole gas)
The airflow at standard condition is converted to field conditions
using the ideal gas law:
Equation 6.104
where:
Tstandard
= air temperature at standard conditions
Tfield
= air temperature at field conditions (°C)
Estimate the
SAE
The SAE is
estimated based on the correlation provided by Praxair. The
expression is valid for SOTE in the range of 86% to 100%. If the
user specified SOTE is less than 80% then the SAE is bounded at
7.11 kg/kWh.
Equation 6.105
where:
SAE = Specific aerator energy
(wire), kgO2/kW-hr
SOTE = Standard oxygen transfer rate (%)
The mechanical power requirement is estimated using the following
expression:
Equation 6.106
where:
Pmechanical
= mechanical power (kW)
SAE
= mechanical aerator oxygen transfer rate (kg O2/kW
h)
CF2
= conversion factor
ηmotor
= motor efficiency
Estimation of Effluent Gas
Composition
The open basin HPO considers the gas-liquid transfer of
O2, CO2, N2, H2 and
CH4 in Mantis2 model. In Mnatis3 model, gas liquid
transfer of N2O is also modeled. The partial pressure of
each gas at the surface of the tank is estimated by assuming a
virtual gas headspace having a volume equivalent to the volume of
the gas holdup in the tank. The virtual head space thus represents
the volume in the tank occupied by the bubbles.
The volume of the
virtual headspace is estimated by using the following
expression:
Equation
6.107
where:
Vheadspace
= volume of headspace (m3)
Vliquid
= volume of liquid in tank (m3)
Φholdup
= gas holdup in tank (-)
The partial pressure of each gas in the headspace is then
integrated using the following derivative equation:
Equation 6.108
where:
gi,o
= partial pressure of i gas, atm
ptotal
= total pressure of the gas in headspace (atm)
rg,i
= gas transfer rate for i gas, mole/m3/d
Vliquid
= volume of liquid, m3
fmolarvolume
= molar volume of gas at standard condition
gi,in
= partial pressure of feed gas, atm
Qgas,
in
= feed gas flow rate, m3/d
Qvent
= outlet gas flow rate, m3/d
In general, the relationship between the growth kinetics and the
temperature may be expressed using an asymmetrical bell shaped
curve. The crest of the curve represents the maximum growth
rate at optimum temperature. The growth rate on either side
of the crest decreases according to some appropriate growth
rate-temperature relationship.
There are a number
of models which may be applied to model the temperature dependent
kinetics for biological reactions. Depending on the structure
of the model, the model may require estimation of model parameters
which may be very different than the traditional temperature
coefficient approach used in the wastewater engineering.
Generally, the Arrhenius equation is used to model the temperature
effect on kinetic parameters in the range of
5oC-35oC. This equation requires one
parameter of temperature coefficient. The values of
temperature coefficients are well researched. Therefore, using any
other model which deviates from this relationship will require
recalculating of model parameters. To model such a curve the
following three parameter model was used.
1.
Estimate the kinetic coefficient at the optimum temperature – The
kinetic parameter values in GPS-X are listed at 20°C; therefore, this value
is translated to a value at the optimum temperature using the
following equations:
Equation 6.109
Equation 6.110
where:
kopt = kinetic
coefficient at Topt, unit
k20 = kinetic
coefficient at 20°C, unit
Topt = optimum
temperature at which the value of kinetic coefficients is
maximum, °C
θ1 =
Arrhenius coefficient for low temperature
θ2
= Arrhenius coefficient for high temperature
2.
Estimate the value of kinetic coefficient at a given temperature T
by using the following equations:
Equation 6.111
Equation 6.112
where:
kT =
kinetic coefficient at T, unit
T = actual
temperature °C
Based on the above relationships the temperature dependency of the
kinetic coefficients can be expressed as shown in Figure
6‑31. For the low temperature range, the default
GPS-X temperature coefficient values were used for all the kinetic
parameters. For the high temperature range temperature
coefficient (θ2), the default values of 1.15 and 1.25 were used for
the growth rate of heterotrophic and autotrophic microorganisms
respectively.
Figure 6‑31 - Variation of Nitrifier Kinetic
Parameter Values with Temperature
The current GPS-X model uses a user-defined temperature in the
biological unit processes. In using these models, it is assumed
that the user has previous knowledge about the expected temperature
in different biological reactors. In this project, it was
decided to estimate the temperature in the biological units based
on the heat balance in the tank considering heat losses and
gains.
A simple heat balance model as proposed by van der Graff
(1976) and later validated by Gillot and
Vanrolleghem (2003) was used to model the temperature
change in a biological tank. The model was appropriately modified
to account for the heat generation in denitrification and
nitrifications reactions. The details of the model are described in
following sections.
The equations
described in Gillot and Vanrolleghem (2003) were
developed for a completely mixed basin under steady state
conditions. These equations were appropriately modified to model
the temperature under dynamic conditions. The completely mixed
hypothesis assumes that the water temperature is uniform over the
basin, and equals the outlet temperature. The energy balance over
the reactor can be expressed with following equation.
Equation 6.113
where:
V =
volume of the reactor, m3
pw
= density of water, kg/m3
cpw =
the specific heat of water, J/kg/°C
Qw =
the wastewater flow rate, m3/s
Twi =
the influent temperature, °C
Two = the
water temperature in reactor, °C
ΔH = enthalpy change due
to heat transfer, J/s
The ΔH term includes the following terms :
1.
Convective and evaporative loss at the surface of the reactor due
to aeration. (Hi)
2.
Heat loss due to conduction across the reactor wall
(Htw)
3.
Heat input due to mechanical and blower energy
(Hp)
4.
Heat input due to biological heat production
(Hb)
The calculation method for each heat transfer term is shown in
Table 6‑8. The present model neglects the heat
transfer due to solar radiation and long wave radiation. The
surface convection and evaporation losses are lumped into the
convection and evaporative heat transfer due to aeration.
Table 6‑8 –
Heat Transfer Terms with Equations for Estimation
Heat Transfer Term
|
Equation Used for
Estimation
|
Comment
|
Hi
|
Ui.A.(Ta-Two)
|
Convective
and evaporative loss during aeration
|
Htw = Htlaw + Htlsw
|
Htlaw
=UlaAla(Two-Ta)
Htlsw
=UlsAls(Two-Ts)
|
Heat transfer
liquid/air contact wall
Heat transfer
liquid/soil contact wall
|
Hp
|
Paer
|
Power input
of surface aerator or blower
|
Hb
= Hbaer+ Hbdn+ Hbaut
|
Hbaer = Huaer*rOU
Hbdn = Hudn*rDN
Hbaut = Hunr*raut
|
Heat
generation from the aerobic oxidation
Heat
generation from denitrification
Heat
generation from nitrification
|
The list of parameters used in the temperature estimation model is
shown in Table 6‑9.
Table 6‑9 –
Parameters for the Temperature Model
Parameter
|
Description
|
Unit
|
Value
|
Ui
|
Heat coefficient
|
W/m2/C
|
Subsurface aeration, Ui =
25
Surface aeration, Ui =
11.4.Paer/V
|
A
|
Surface area of reactor
|
m2
|
Variable
|
V
|
Volume of the reactor
|
m3
|
Variable
|
Ta
|
Air temperature
|
C
|
Variable
|
Two
|
Water temperature in reactor
|
C
|
Variable
|
Ula
|
Heat transfer coefficient for liquid/air wall
|
W/m2/C
|
1.0
(default)
|
Ala
|
Liquid-air contact wall area
|
m2
|
Variable
|
Uls
|
Heat transfer coefficient for liquid/air wall
|
W/m2/C
|
1.0
(default)
|
Ts
|
Soil temperature
|
C
|
Variable
|
Als
|
Liquid-soil contact wall area
|
m2
|
Variable
|
Paer
|
Total aerator power
|
W
|
Variable
|
Huaer
|
Unit heat production during aerobic reaction
|
J/g O2 consumed
|
13985.0
(default)
|
Hudn
|
Unit heat production during denitrification
|
J/g NO3-N consumed
|
32000.0
(default)
|
Huaut
|
Unit heat production during nitrification
|
J/g NH4-N consumed
|
25000.0 (default)
|
rou
|
Oxygen uptake
rate
|
g
O2 consumed/s
|
Estimated by
model
|
rdn
|
denitrification rate
|
g
NO3-N consumed/s
|
Estimated by
model
|
raut
|
Ammonia
oxidation rate
|
g
NH3-N oxidized/s
|
Estimated by
model
|
All the heat transfer terms are estimated, and the heat balance
equation is solved is solved both for steady state and dynamic
conditions. An iterative Newton-Raphson method is used to
solve for temperature at steady state.
The powdered
activated carbon (PAC) object is a normal CSTR object that has been
enhanced with powdered activated carbon addition. The carbon
sorbs one or more components using a competitive sorption
isoterm. The sorbed components are then coverted to
particulate inerts before moving downstream. The version of
the PAC model is different in the IP libraries, where 3
user-defined components can be added to the existing sorbable set
of state variables, and used as toxic inhibitors to biological
activity.
Note the Powdered
Activared Carbon object is only available in the GPS-X legacy
libraries (cnlib, cniplib, cnplib, cnpiplib)
Sorption Isotherms
In the CN and CNP
libraries, the only components being sorbed from solution onto the
PAC are ss (soluble substrate) and si (soluble
inerts). In this two-component system, the sorption rates
are:
Equation 6.114
Equation 6.115
where:
adsrateSI
= adsorption
rate of SI (gCOD/m3/d)
adsrateSS
= adsorption rate of SS (gCOD/m3/d)
MASI
= maximum sorption capacity for SI (gCOD/gPAC)
MASS
= maximum sorption capacity for SS (gCOD/gPAC)
KLSI
= SI sorption half-sat coefficient (gCOD/m3)
KLSS
= SS sorption half-sat coefficient (gCOD/m3)
XPAC
= PAC concentration in liquid (gPAC/m3)
RADS
= adsorption rate (1/d)
The PAC model calculates the amount of “fresh” PAC (PAC that is
available to sorb components), and “used” PAC (PAC that has
components sorbed onto it, and is not available to sorb more), and
the fraction of total PAC that is “fresh” and available to
sorb.
The model calculates the concentrations of individual components
that are bound to PAC. For PAC-bound soluble substrate, the
biomass will grow at an enhanced rate, due to the presence of a
concentrated substrate source. This enhancement of growth is
controlled by the bioregeneration factor, by which all of
the biological growth rates (that grow on ss) are multiplied.
The consumption of PAC-bound ss frees up the “used” PAC to
become “fresh” PAC again.
In the CNIP and
CNPIP libraries, the above-described functionality is enhanced by
the addition of user-defined toxic components and their sorption
properties. Users can use the IP library versions of the PAC
object to investigate the relationship between toxic inhibition and
PAC addition to sorb the toxics.
Toxic Inhibition
The growth rates
for biomass (both heterotrophic and autotrophic) in the Mantis
biological model is appended with a switching function for the
presence of toxic inhibitors. The user must identify which
components are the toxic components, and specify the relative
inhibition of each component. The Toxic Inhibition
menu contains the parameters used in the inhibition switching
function. Figure 6‑32 shows the Toxic
Inhibition menu, with sza (IP state variable “soluble
component ‘a’”) filled in as Toxic Component #1.
Figure 6‑32
– Toxic Inhibition Menu
Note, the user must specify the cryptic variable name without
labels. The user may specify any state variable as a toxic
component, including regular states (sno, snh, xnd, etc.)
and any of the IP state variables (sza, szb,
xza, xzb, etc.).
The toxic component
# 1 inhibition constant is a half-saturation coefficient that
represents the concentration at which the biological growth rate
will be half of its normal value.
Sorption of Toxic
Components
The sorption parameters for the toxic components are found in the
PAC Addition menu, in the More… button. These
components are added to the ss and si sorption
isotherms described above, to make a 5-component competitive
sorption system. The sorption rates for a five-component
system are all on the form.
Equation 6.116
where:
var1, var2,
var3
= toxic component concentrations (g/m3)
MAvar
= maximum adsorption capacity (g/gPAC)
KLvar
= sorption half-saturation coeff (g/m3)
The multi-component competitive sorption model will favour sorption
of those components with the lowest sorption half-saturation
coefficient.
CHAPTER
7
This chapter
describes the aerobic attached-growth models available in GPS-X.
These models are associated with the trickling filter, rotating
biological contactor (RBC), submerged biological contactor
(SBC), simple/advanced biological aerated filter
(BAF) and hybrid (IFAS or MBBR) objects. The
biofilm model is also used in the anoxic denitrification filter
object in the Tertiary Treatment process model group.
The major
difference between these models and the suspended-growth models
described in Chapter 6 (page 95) is the inclusion of the diffusion
process in the biofilm. Although diffusion occurs in
suspended-growth systems (substrate and oxygen must diffuse into
the activated sludge floc), it is usually neglected since the
rate limiting steps are the biological reactions; however, in
attached-growth processes, the diffusion and biological reactions
must be considered, increasing the complexity of the models.
The trickling
filter, RBC, SBC and hybrid processes are similar. They provide
biofilm biological treatment, with excess biofilm sloughing
separated in a clarifier object. The SBC and BAFs are flooded
processes, with mechanical supplemental aeration.
The BAFs differ
from the other processes in that it:
·
is a combined biological and solids separation process, and
·
has a mechanical process for controlling excess biomass growth and
captured solids (i.e. backwash)
The advanced BAF
model is a combination of an attached-growth model, and a
filtration model. The filtration model is described in the
One-Dimensional Models section of CHAPTER 8.
Introduction
The trickling filter model is available in all the libraries, using
the same biological reactions found in the suspended-growth models
in the appropriate library. The model can predict the extent of
carbon and nitrogen removal (by uptake or oxidation) and
denitrification, and phosphorus uptake and release. This model
incorporates the growth kinetics and transport processes for the
corresponding state variables. The profiles of the various
components through the biofilm are modelled so that different
environments (aerobic, anoxic and anaerobic) can exist within the
biofilm.
To reduce the complexity of the model, some assumptions are
necessary. The limitations of this model concern the hydraulics of
the trickling filter and the biofilm itself. The model assumes that
the flow rate and solids loading to the filter can always be
processed; that is, clogging and head losses through the filter are
not modelled. Also the maximum thickness of the biofilm is not
calculated, rather the user specifies it. This assumption was made
because there are little or no data available for
calibration/verification of the maximum film thickness
calculations. It is assumed that there is equal flow distribution
over the entire surface area of the trickling filter and the media
inside the trickling filter. The effect of the rotation speed of
the rotary distributor is neglected.
The dimensions of
the trickling filter model are larger than the suspended-growth
models since the state variables are now modelled through the film
as well as down through the trickling filter. The suspended-growth
models only considered the state variables along the reactor
(1-dimensional). The additional dimension in the biofilm has some
impact on the simulation speed; therefore, more time is required
when using this model. Improvements to the speed of this model have
been made by integrating the state variables with different
frequencies. For example, the particulate components were found to
change more slowly than the soluble components since they are
diffusing through the biofilm. Therefore, the integration of the
soluble components was handled differently from the particulate
components (see Figure 7‑3).
Conceptual Model
The trickling
filter is divided into ‘n’ horizontal sections (default is six
sections) each representing a cross-section of the trickling filter
at a different depth. The transfer of the state variables
between each of these horizontal sections through the liquid film
is through liquid flow. The biofilm in each of these
horizontal sections is modelled as a number of layers (default is
one layer for the liquid film on top of five layers for the
biofilm). The transfer of soluble state variables between
each of these layers is by diffusion only. Particulate variables
have a certain physical volume associated with them and can be
displaced into the neighbouring layer by growth processes.
Each layer of the
biofilm is modelled as a CSTR with the same biological reactions as
the suspended-growth biological model (See Appendix A for the
mantis model). Attachment and detachment coefficients
are used to provide for a means of transfer of particulate
components between the biofilm surface and the liquid film.
This conceptualization is shown in Figure 7‑1.
The concentration of each particulate state variable is converted
to volume based on the dry material content of biofilm and
the density of biofilm (both user inputs). When the
volume of each layer is filled (based on maximum biofilm
thickness and number of biofilm layers) the next layer
begins to fill. When the film thickness approaches the specified
maximum, increasing detachment of biofilm will occur.
Figure 7‑1 -
Conceptual Diagram of the Tricking Filter Model
Mathematical Model
The mathematical
equations used within each layer of the biofilm are provided in the
corresponding Model matrix (for example, see Appendix A for the
mantis model). The equation used for the diffusion of the
state variables from the bulk liquid into the biofilm is provided
below. The diffusion through the biofilm is described by Fick's
second law and supplemented with biological reactions:
Equation 7.1
where:
Aa
= surface area of biofilm through which transport is occurring
(m2)
δL
= thickness of attached liquid layer (m)
SjL
= substrate conc. in liquid film horizontal section
(mg/L)
t
= time (days)
S
jBLi = substrate
concentration at biofilm liquid interface section j
(mg/L)
So
= saturated liquid-film substrate concentration (mg/L)
QL
= volumetric flow rate of attached liquid layer (L/d)
K M
= mass transfer coefficient from liquid to biofilm (m/d)
K ML
= oxygen transfer coefficient from air to liquid film
(m/d)
Equation 7.2
where:
A
= surface area of attached microorganisms (m
2)
DS
= state variable
diffusion coefficient (m2/d)
QB
= volumetric flow rate of attached biofilm layer (L/d)
Rs
= substrate utilization rate (mg/L/d)
S
= state variable concentration in layer (mg/L)
S
jB
= state variable concentration in attached biofilm layer j
(mg/L)
t
= time (days)
y
= thickness of biofilm layer (m)
dL
= attached biofilm thickness in layer (m)
Model Parameters
This section of the
chapter discusses the various model parameters and inputs that the
user would encounter when using this model. The examples and
discussion below pertain to the mantis model in the CN
library.
Physical
These menu items
are found under the Parameters sub-menu item Physical
and contain both real physical dimensions to describe the actual
trickling filter being modelled, as well as model dimensions which
allow the user to specify how the physical system will be modelled.
There are two items under the heading Speed, which allow the
user to optimize the simulation speed of this model by changing the
frequency of integration for the soluble components.
As seen in
Figure 7‑2, the real dimensions of the trickling
filter require inputs such as: the filter bed depth,
filter bed surface and specific surface of media.
Together, these three parameters will provide an estimation of the
total biofilm surface area in the model. The liquid retention
time in filter refers to the retention time of the liquid
flowing past the biofilm, the maximum attached liquid film
thicknessrefers to the liquid film which is not moving past the
biofilm due to friction (no-slip layer) and the maximum biofilm
thickness will be reached only with infinite sloughing. The
density of biofilm and dry material content of
biofilmare used to convert the concentration of each state
variable to a volume measurement.
Figure 7‑2 -
Physical Dimensions of the Tricking Filter
The model dimensions are fixed, that is, the number of layers in
the biofilm is six (one liquid layer and five biofilm layers).
Also, the number of horizontal and vertical sections in the model
is fixed at six and one respectively.
The
next two items in this form, under the heading Speed,
concern the integration of the soluble state variables. Since these
variables are diffusing through the biofilm, and consequently
change rapidly when compared to the particulate state variables,
which are changing only due to their growth rates, they tend to
dominate the numerical solver. To increase the speed of simulation,
the soluble states can be integrated less frequently without loss
of accuracy. When the variables are scheduled to be integrated
their derivatives are calculated from equations
(Equation 7.1) and
(Equation 7.2)
shown above. Otherwise, the derivatives are set to zero. There are
two variables used to specify how frequently the soluble states are
integrated and the duration for integration or how long the states
are integrated for each period: the soluble integration
period and soluble integration length. These concepts
are shown in Figure 7‑3 with the results on ammonia
shown in Figure 7‑4.
Figure 7‑3 -
Integration of Soluble Components
Figure 7‑4 –
Integration
Transport
The parameters
shown in Figure 7‑5 are used for the transport model
(diffusion) of the various components through the biofilm. Since
there are insufficient data in the literature concerning the
diffusion of various components through a biofilm, the values used
for the diffusion through water are reduced by a constant fraction,
shown as reduction in diffusion in biofilm. The default
diffusion coefficients shown for water are used for the diffusion
of components from the liquid layer to the first layer (outside) of
the biofilm. The detachment rate and attachment rate
are used for calculating the sloughing rate and particulate
components attachment to the biofilm, respectively.
Figure 7‑5 -
Mass Transport Parameters
The physical
dimensions of the tricking filter show more physical components
associated with oxygen. The oxygen mass transfer coefficient is
calculated from the physical conditions within the filter
(thickness of biofilm and diffusion rate of oxygen) and is affected
by temperature and the saturated dissolved oxygen concentration.
These values are either input in the general data entry area or
specifically set for each object as shown in
Figure
7‑6.
Figure 7‑6 -
Physical Dimensions of the Trickling Filter
Stoichiometric and Kinetic
Parameters
The parameters
shown in these menus refer to the biological reactions that are
described in CHAPTER 6 .
Output Variables
In addition to the
standard effluent parameters, there are a number of fixed film
specific variables that can be displayed. Figure 7‑7
shows the variables available for display for the trickling
filter.
This includes the
biofilm thickness for each horizontal section from the top to the
bottom of the filter.
Biofilm Profiles
This provides all
the state variables plus the suspended solids composite variable
for each horizontal section of the trickling filter (i.e. six
profiles for six horizontal sections). The first layer in the
profile is the liquid film, followed by the biofilm layers (i.e.
liquid and maximum of five biofilm layers).
Figure 7‑7 -
Tricking Filter Output Variables
This provides the
liquid film state variables for each horizontal section of the
trickling filter, from top (section 1) to bottom (section 6).
This is used for plotting both the filter horizontal sections and
biofilm layers variables using the 3D bar chart or grayscale
outputs. Examples are shown in Figure
7‑8. From left to right are the liquid layer
and five biofilm layers. In the 3D bar chart, the top of the
filter is located in the foreground, and the bottom of the filter
in the background. In the greyscale graph, the six
horizontal trickling filter horizontal sections are shown from top
to bottom (i.e. top layer of the filter is shown at the top of the
graph).
Figure 7‑8 -
2D Variables
Introduction
The rotating
biological contractor (RBC) model is available in all the
libraries, using the same biological reactions found in the
suspended-growth models in the appropriate library. The model can
predict the extent of carbon and nitrogen removal (by uptake or
oxidation) and denitrification, as well as phosphorus uptake and
release. This model incorporates the growth kinetics and transport
processes for the corresponding state variables. The profiles of
the various components through the biofilm are modelled so that
different environments (aerobic, anoxic and anaerobic) can exist
within the biofilm.
To reduce the complexity of the model, some
assumptions are necessary. The limitations of this model concern
the hydraulics of the rotating biological contactor and the biofilm
itself. The model assumes that the flow rate and solids loading to
the RBC can always be processed; that is, clogging and head loss
are not modelled. Also the maximum thickness of the biofilm is not
calculated; rather it is specified by the user. This assumption was
primarily made because there are little or no data available for
calibration/verification of the maximum film thickness
calculations. The effect of the rotation speed and direction of the
RBC and its impact on media sloughing and aeration requirements is
neglected.
Similar to the
trickling filter model, the RBC is more complex than the
suspended-growth models since the state variables are modelled
through the biofilm as well as through various RBC stages.
Conceptual Model
The rotating
biological contactor is divided into ‘n’ stages (default is 1
stage) each representing a baffled RBC system. The transfer of the
state variables between each of these stages is through the liquid
flow. The biofilm in each stage is modelled as a number of layers
(default is one layer as the liquid film on top of five layers as
the biofilm). The transfer of soluble state variables between each
of these layers is by diffusion only. Particulate variables have a
certain physical volume associated with them and can be displaced
into the neighbouring layer by growth processes. Each layer
of the biofilm is modelled as a CSTR with the same biological
reactions as the suspended-growth biological model (See Appendix A
for the mantis model). Attachment and detachment
coefficients are used to provide for a means of transfer of
particulate components between the biofilm surface and the liquid
film.
This
conceptualization is shown in Figure 7‑9. The
concentration of each particulate state variable is converted to
volume based on the dry material content of biofilm and its
density, both input by the user. When the volume of each biofilm
layer is filled (based on maximum biofilm thickness and number of
biofilm layers) the next layer begins to fill. When the biofilm
thickness starts to approach the specified maximum, increasing
detachment of biofilm will occur.
Figure 7‑9 –
Conceptual Diagram of the RBC Model
Mathematical Model
The mathematical
equations used within each layer of the biofilm are provided in the
corresponding Model matrix (For example, see Appendix A for the
mantis model). The equation used for the diffusion of the
state variables from the bulk liquid into the biofilm is provided
in the trickling filter section.
Model Parameters
The mathematical
equations used within each layer of the biofilm are provided in the
corresponding Model matrix (For example, see Appendix A for the
mantis model). The equation used for the diffusion of the
state variables from the bulk liquid into the biofilm is provided
in the trickling filter section.
Physical
These menu items
are found under the Parameters sub-menu item Physical
and contain both real physical dimensions to describe the actual
rotating biological contactor modelled, as well as model dimensions
which allow the user to specify how the physical system will be
modelled. There are two items under the heading Speed, which
allow the user to optimize the simulation speed of this model by
changing the frequency of integration for the soluble components.
As seen in Figure 7‑10, the real dimensions of
the rotating biological contactor require inputs such as the rbc
liquid volume, the rbc media volume and the specific
surface area of media. Together, these three parameters provide
an estimation of the total biofilm surface area in the model. The
submerged fraction of the biofilm refers to the percent of
the RBC submerged at any given time, the maximum attached liquid
film thickness refers to the liquid film which is considered to
be associated with the biofilm due to friction (no-slip layer), and
the maximum biofilm thickness will be reached only with
infinite sloughing. The density of biofilm and dry
material content of biofilm are used to convert the
concentration of each state variable to a volume measurement. The
model dimensions include the number of RBC tanks in
series or stages. The model assumes each tank or stage is of
equal size. If the stages are of unequal size (different media
density and/or parallel feeding), the user should tie individual
RBC objects together to simulate the process.
Figure 7‑10
- Physical Dimensions of the RBC
The next two items
in this form, under the heading Speed, concern the
integration of the soluble components. Since these components are
diffusing through the biofilm, and consequently change rapidly when
compared to the particulate components, which are changing only due
to their growth rates, they tend to dominate the numerical solver.
To increase the speed of simulation, it was found that these
components can be integrated less frequently without loss of
accuracy. When the soluble variables are scheduled to be
integrated, their derivatives are calculated from equations
(Equation 7.1) and (Equation 7.2).
Otherwise, the derivatives are set to zero. There are two variables
which are used to specify how frequently the states are integrated:
the soluble integration period, and the duration for
integration or how long the states are integrated for each period
(soluble integration length). These concepts are shown in
Figure 7‑3, with the results for ammonia shown in
Figure 7‑4.
Figure
7‑11 shows more physical components associated with oxygen
solubility. The oxygen mass transfer coefficient is calculated from
the physical conditions within the filter (thickness of biofilm and
diffusion rate of oxygen) and is affected by temperature and the
saturated dissolved oxygen concentration. These values are either
input in the general data entry area or specifically set for each
object.
Figure 7‑11
– Physical Dimensions of the RBC (More…)
Mass Transport
The parameters
shown in Figure 7‑5 are used by the RBC for the
transport model (diffusion) of the various components through the
biofilm. Since there are insufficient data in the literature
concerning the diffusion of various components through a biofilm,
the values used for the diffusion through water are reduced by a
constant fraction, shown as reduction in diffusion in biofilm. The
default diffusion coefficients shown for water are used for the
diffusion of components from the liquid layer to the first layer
(outside) of the biofilm. The rate of detachment and attachment is
used for calculating the sloughing rate and particulate components
attachment to the biofilm.
Stoichiometric and Kinetic
Parameters
The parameters
shown in these menus refer to the biological reactions that are
described in CHAPTER 6.
Output Variables
In addition to the
standard effluent parameters, there are a number of fixed film
specific variables that can be displayed. These are accessed
through the object’s Output Variables menu item.
·
RBC Variables (see Trickling Filter Variables
section in this chapter)
·
Liquid Concentrations (see Liquid Film
Concentrations section in this chapter)
·
2D Variables (see 2D Variables section in this
chapter)
Introduction
The submerged
biological contactor (SBC) model is a modification of the RBC model
for units that are air driven or are provided with supplemental
aeration. Units with supplemental aeration tend to be more
submerged than conventional RBCs (conventional RBCs are generally
40 percent submerged) providing additional contact of the media
with the influent, since the unit is not dependent on ambient air
for oxygen.
The SBC model discussed
herein is available in all the libraries, using the same biological
reactions found in the suspended-growth models in the appropriate
library. The model can predict the extent of carbon and nitrogen
removal (by uptake or oxidation) and denitrification, as well as
phosphorus uptake and release (in the carbon-nitrogen-phosphorus
library). This model incorporates the growth kinetics and transport
processes for the corresponding state variables. The profiles of
the various components through the biofilm are modelled so that
different environments (aerobic, anoxic and anaerobic) can exist
within the biofilm.
To reduce the complexity of the model, some assumptions are
necessary. The limitations of this model concern the hydraulics of
the submerged biological contactor and the biofilm itself. The
model assumes that the flow rate and solids loading to the SBC can
always be processed; that is, clogging and head loss is not
modelled. The maximum thickness of the biofilm is not calculated;
rather it is specified by the user. This assumption was made
because there are little or no data available for
calibration/verification of the maximum film thickness
calculations. It is assumed that there is equal flow distribution
over the entire surface area of the SBC and the media inside the
SBC. The effect of the rotation speed or direction of the SBC and
its impact on media sloughing is neglected. The oxygen diffused
into the biofilm when the media is in the air is neglected, since
the SBC media is generally submerged more than 80 percent of the
time; therefore, the SBC is modelled as completely submerged.
Similar to the trickling filter
model and the RBC, the SBC is more complex than the
suspended-growth models since the state variables are now modelled
through the film and through various SBC stages (or shafts).
Conceptual Model
The submerged
biological contactor is divided into a number of stages (default is
2 stages) each representing a baffled SBC shaft. The transfer of
the state variables between each of these stages is through the
liquid flow. The biofilm in each stage is modelled as a number of
layers (default is one layer as the liquid film on top of five
layers as the biofilm). The transfer of soluble state variables
between each of these layers is by diffusion. Particulate variables
have a certain physical volume associated with them and can be
displaced into the neighbouring layer by growth processes. Each
layer of the biofilm is modelled as a CSTR with the same biological
reactions as the suspended-growth biological model (See Appendix A
for the mantis model). Attachment and detachment
coefficients are used to provide for a means of transfer of
particulate components between the biofilm surface and the liquid
film.
For the SBC the
process is the same as the RBC (conceptually shown in Figure
7‑9). The concentration of each particulate state variable
is converted to volume based on the dry material content of biofilm
and its density input by the user. When the volume of each layer is
filled (based on maximum biofilm thickness and number of biofilm
layers) the next layer begins to fill. When the film thickness
starts to approach the specified maximum, increasing detachment of
biofilm will occur.
Mathematical Model
The mathematical
equations used within each layer of the biofilm are provided in the
corresponding Model matrix (for example, see Appendix A for the
mantis model). The equation used for the diffusion of the
state variables from the bulk liquid into the biofilm is provided
in the trickling filter section.
Model Parameters
This section of the
chapter discusses the model parameters and inputs that the user
will encounter using this model, in particular those different from
the other fixed film models. The example and discussion below
pertains to the mantis model in the Carbon - Nitrogen (CN)
library.
Physical
These menu items are found under Parameters > Physical,
and contain both real physical dimensions to describe the actual
submerged biological contactor being modelled, and model dimensions
which allow the user to specify how the physical system will be
modelled. there are two items under the heading Speed, which
allow the user to optimize the simulation speed of this model by
changing the frequency of integration for the soluble
components.
As seen in Figure
7‑12, the real dimensions of the submerged biological
contactor require inputs such as:
1.
tanks in series
2.
volume set up method
3.
SBC liquid volume in tanks
4.
SBC media volume in tanks
5.
SBC total volume
6.
SBC total media volume
7.
volume fractions
8.
Specific surface of media (which will provide an estimation of the
total biofilm surface area in the model.)
Based on the volume set up method (input 2), the user either
specifies the individual volumes per tank (inputs 3 and 4), or the
volume fractions based on the total volumes (inputs 5, 6 and 7).
The latter two items are used to convert the concentration of each
state variable to a volume measurement. Model dimensions
include the number of SBC tanks in series or stages.
Figure 7‑12
– Physical Dimension of the SBC
Biofilm
characteristics include:
1.
Maximum attached liquid film thickness, which refers to the
liquid film that is considered to be associated with the biofilm
due to friction (no-slip layer).
2.
Maximum biofilm thickness, which will be reached with
infinite sloughing
3.
Density of biofilm
4.
Dry material content of biofilm
The next two items in this form under the heading Speed,
concern the integration of the soluble components. Since these
components are diffusing through the biofilm, and consequently
change rapidly when compared to the particulate components, which
are changing only due to their growth rates, they dominate the
numerical solver. To increase the speed of simulation, it was found
that these components could be integrated less frequently without
loss of accuracy. When the soluble variables are scheduled to be
integrated their derivatives are calculated as normal. Otherwise,
the derivatives are set to zero. There are two variables which are
used to specify how frequently the states are integrated
(Soluble integration period) and the duration for
integration or how long the states are integrated for each period
(Soluble integration length). These concepts are shown in
Figure 7‑3, with the results for ammonia shown in
Figure 7‑4.
The oxygen mass transfer coefficient is calculated from the
physical conditions within the filter (thickness of biofilm and
diffusion rate of oxygen) and is affected by temperature and the
saturated dissolved oxygen concentration. These values are input in
the general data entry area or specifically set for each object as
shown in Figure 7‑13.
Figure 7‑13
- Physical Dimensions of the SBC (More...)
Operational
The operational
parameters are similar to the activated sludge model. Like
the activated sludge model, the aeration method can be chosen as
either “Enter KLa” or
“Enter Airflow”. No mechanical method is provided since
this method is not appropriate for the SBC.
Mass Transport
The parameters
shown in Figure 7‑5 are used by the SBC for the
transport model (diffusion) of the various components through the
biofilm. Since there are insufficient data in the literature
concerning the diffusion of various components through a biofilm,
the values used for the diffusion through water are reduced by a
constant fraction, shown as reduction in diffusion in biofilm. The
default diffusion coefficients shown for water are used for the
diffusion of components from the liquid layer to the first layer
(outside) of the biofilm. The rate of detachment and attachment is
used for calculating the sloughing rate and particulate components
attachment to the biofilm.
Stoichiometric and Kinetic
Parameters
The parameters
shown in these menus refer to the biological reactions that are
described in CHAPTER 6.
Output Variables
In addition to the
standard effluent parameters, there are a number of fixed film
specific variables that can be displayed. They are the same as the
trickling filter. These are accessed through the following
subheadings:
·
SBC Variables (see Trickling Filter Variables
section in this chapter)
·
Liquid Concentrations (see Liquid Film
Concentrations section in this chapter)
·
2D Variables (see 2D Variables section in this
chapter)
Conceptual Model
The simple BAF model combines the 1-D biofilm model used in the
trickling filter with an aeration model and simple
solids-separation model. It is similar in construct to the
denitrification filter, with the addition of aeration. The
Simple BAF model is designed to be a less-sophisticated, but
easier-to-use alternative to the advanced BAF model. The
Simple BAF model will solve to steady-state using the GPS-X
steady-state solver (whereas the Advanced BAF model cannot, and
requires a dynamic simulation to come to equilibrium).
The simple BAF uses
a series of horizontal layers (6 by default) to represent plug-flow
through media. The filter is fed from the bottom and effluent
is taken from the top.
Figure 7‑14 - Simple BAF Model
Configuration
Oxygen solubility
is calculated separately for each layer, allowing for depth effects
to be reflected in oxygen saturation concentration.
Hydraulics
Influent flow
enters the unit through the lower-left connection point, and exits
from the upper right-hand connection point. Solids that are
periodically backwashed from the unit are converted to a continuous
flow that exits from the lower-right connection point.
Physical Parameters
The physical
parameters menu for the Simple BAF model contains parameters for
the description of the unit, including the characteristics of the
media. Figure 7‑15 shows the physical
parameters menu. The definitions of the various biofilm
parameters (e.g. maximum biofilm thickness, etc.) are
identical to those described in the section on the Trickling Filter
model. The default parameter values for the media
characteristics are set for typical BAF media.
Figure 7‑15
- Simple BAF Model Operational Parameters Menu
Operational
Parameters
The operational parameters menu contains settings for the aeration
and backwashing of the filter.
The solids capture fraction is used to determine the mass of
solids captured on the filter. These solids are then removed
via the
The backwash flow rate and backwash duration per 24-hr
period are used (along with the calculated captured solids) to
calculate a mass flow of backwashed solids. The backwashed
flow is converted from an intermittent event to a continuous flow
of equivalent mass solids. This allows for the model to be
solved using the steady-state solver.
Details of the aeration setup can be found in the More...
button under in the Aeration Setup section. (Figure
7‑16)
Figure 7‑16
- Simple BAF Model Operational Parameters Menu (More...)
The remaining menus
are equivalent to those found in the Trickling Filter Model.
The advanced
biological aerated filter (BAF) model is a robust,
mechanistically-based model. It uses the same biological
reactions found in the suspended-growth models discussed in Chapter
6.
Conceptual Model
The advanced BAF
model consists of four major components: Hydraulics and Filter
Operation, Filtration, Biological Reactions, and a Biofilm. They
are described in the following sections.
Hydraulics and Filter
Operation
The model needs to be able to describe the complex operation of the
bio filter stages each consisting of one or more units in different
operating modes.
In the simplest implementation (number of units is 1); the
model simulates the behavior of a single BAF unit. This unit can be
in filtration, standby, backwash, or
flush mode. In filtration, standby and
flush mode, the filter is represented hydraulically as a
tank consisting of a certain number of horizontal sections (6 in
the default case). The actual filtration bed is preceded by a mixed
tank without filter material to describe the dilution effects of
the volume of water under the filter bed. A similar mixed tank is
added to account for the liquid on top of the filter media. In
backwash mode the horizontal sections of the filter are
combined and converted to one mixed tank. The model assumes that
during backwash the filter media is ideally mixed.
The influent loading to the filter determines if the filter is in
filtration, standby, backwash, or flush mode. In filtration and
flushed modes, flow is entering the filter through the input stream
and coming out through the output stream. In standby mode, there is
no flow through the filter. In backwash mode, flow is entering the
filter through the backwash input stream and coming out through the
backwash output stream.
A complex bio
filter plant could be represented by the proper number of
individual filter units placed on the drawing board to describe the
changing conditions in the plant and forecast effluent quality. Due
to the level of complexity within one filter unit and the typical
number of units in the plant this approach is not feasible even
with substantial computing capacity. Since the difference between
individual units operating in the same mode is not drastic, a
simplified operation mode provides a good approximation of
operating and effluent conditions with substantially less
overhead.
For this purpose
the hydraulics of the filter (which now can be thought of as a
series of units) is described in three different ways. A certain
fraction of the total number of units is in filtration mode,
i.e. layered with influent flowing through them. Another fraction
is in standby mode, with light aeration and no influent
loading. A third fraction is in backwash mode. The sum of
the three fractions adds up to the total number of units. At times
any one of these fractions may be completely missing, i.e. the
number of units in backwash or standby mode can be zero. In GPS-X,
the individual fractions are specified in two ways:
1.
As constants – This allows a simple run when the number of
units in standby does not change. When a particular backwash
criterion is reached, a certain number (user-specified) of units in
a filtration mode are backwashed. This mode can be used if an
actual operation of a filter plant is recorded and has to be
replayed in GPS-X. The number of units can be read in through
the GPS-X input file facility, and the original loading conditions
recreated.
2.
As volume fractions which vary according to a target load on the
filter component – This way the ration of the operating filter
volume, and the standby volume changes continuously according to
changes in influent load. This operation mimics a "constant
loading" operational policy, although the loading is completely
constant on the filter volume because the limitation created by the
individual bio filter unit volumes are ignored.
During the
simulation when volume fractions in different operating modes
change, the model correctly keeps track of mass balances by
recalculating the mass of all model components currently in that
element. As an example consider the event of one standby unit
coming on-line due to increasing load on the filter. The other
elements in the filter, which have been in operation, may contain
much higher active biomass than one, which has been on standby for
a period of time. When the standby component comes on-line, a
volume weighted average biomass concentration is calculated for the
new, increased filter volume for all horizontal sections and all
biofilm layers, and integration continues using the new
conditions.
Filtration Component
The modified Iwasaki equation (Horner et al.,
1986) was used to calculate the filtration rate, while head
loss is calculated according to the Kozeny equation. In the
filtration mode, the number of horizontal sections is determined by
the plug flow characteristics of the BAF bed (6 by default). This
component of the model is used in predicting solids capture,
effluent suspended solids and other constituents and backwash
quality. In backwash mode, the backwashed fraction of the bed is
treated as an ideally mixed tank; biofilm layers are retained.
The filtration element of the
model is available without the biological reactions in the current
sand-filter model in GPS-X described in One-Dimensional
Model section of Chapter 9.
Biological Reactions
Component
The BAF models use
the same biological reactions found in the suspended-growth models
in the appropriate library. The mantis model has been
successfully used in several biofilm configurations by Hydromantis,
and is able to predict ammonia and BOD profiles in the reactor and
in the biofilm. For more specific information on biological models,
please consult CHAPTER 6.
Biofilm Component
The existing
biofilm model in GPS-X is based on Spengel and Dzombak
(1992). This model was adapted to the simple and advanced
BAF configurations. The model handles soluble material diffusion,
biofilm growth, and particulate attachment and detachment. Details
are described in the Trickling Filter Model section
of this chapter.
Mathematical Model
The mathematical
equations used within each layer of the biofilm are provided in the
corresponding Model matrix (see Appendix A for the mantis
model). The equations used for the diffusion of the state variables
from the bulk liquid into the biofilm are provided in the
Trickling Filter Model section of this chapter. The
equations used for the filtration component are provided in the
One-Dimensional Model section of CHAPTER
9.
Model Parameters
This section
discusses the various model parameters and inputs that the user
will encounter when using this model.
Physical
The physical parameters are found under the Input Parameters
sub-menu item Physical. It contains physical dimensions to
describe the actual BAF being modelled and model dimensions, which
allow the user to specify how the physical system will be modelled.
There are three items under the heading Speed, which
allow the user to optimize the simulation speed of this model.
As seen in Figure
7‑17, the unit dimensions of the BAF require inputs
such as the single filter bed surface area, the total
filter bed depth from support, the media fill (empty bed
depth) (the difference between the last 2 inputs giving the
water height above the media), and the water height below
support. The number of units makes the simplified
multiple units operation possible.
The next section of this
form, media, includes inputs to characterize the
media being used in the filter: Specific surface of media
together with the filter bed depth and surface area provides an
estimation of the total biofilm surface area in the model;
Equivalent particle diameter, where multi-media filters can
be described; Clean bed porosity (void space); and
ultimate bulk biofilm volume, the maximum space the biofilm
can take up before completely clogging the filter.
The next two items under the Biofilm heading, density
of biofilm and dry material content of
biofilm, are used to convert the concentration of each
state variable to a volume measurement. The Model Dimensions
section includes the number of sections in filter; the model
assumes each horizontal section is of equal size.
Figure 7‑17 shows physical components associated with
oxygen solubility and model speed. The oxygen mass transfer
coefficient is calculated from the physical conditions within the
filter (thickness of biofilm and diffusion rate of oxygen) and is
affected by the liquid and air temperatures, and the oxygen
fraction in air (See CHAPTER 6). These values are
input in the general data entry area or specifically set for each
object as shown in this screen. The first two items under the
heading Speed concern the integration of the soluble
components, which is described in the Trickling Filter
Model section of this chapter. The third item, calculate
DO in liquid, is used to improve the speed of simulation if the
BAF installation maintains a relatively high level of DO in the
liquid (close to saturation). By setting this parameter to
OFF, a constant DO level is maintained in the liquid and
integration of the DO state variable is bypassed, thereby speeding
up the simulation. The constant DO level can be adjusted in the
initial conditions form (Process Data >
Initialization > initial concentrations).
Figure 7‑17
- Advanced BAF Physical Parameters
Operational
In the first part
of the Operationalmenu item, aeration constants are
available similar to all aerated biological units. Specific to the
advanced BAF model is the ability to specify a different
KLa (or air flow, if the aeration method is set to
diffused) for each of the filter operational states (active,
standby, flushed, or backwashed). These parameters are shown in
Figure 7‑18
Figure 7‑18
- BAF Operational Parameters
The entry under the
Cycles heading, backwash operation, is used to set up
the method by which a backwash operation will be initiated when
operating the filter(s). The backwashoperation can
be time-based, head loss-based, effluent quality based, or
manual. The filter can be operated in single unit
mode (the whole volume will be in filter, backwash, standby or
flushed mode), or multiple unit operation. In multiple unit
operation (if the number of units is greater than one), the
number of units in filter mode can be specified, and the
remaining units are in standby mode. The number of
simultaneously backwashed units will be taken out of the
filtration mode and replaced with standby units when a backwash
operation is initiated if the constant filter loading rate
controller is on. The target loading rate on the filter
fraction can be specified and thus the volume of filter in
filtration mode will vary to maintain the target loading rate.
This form contains further information about the backwash
operation. Backwash can be initiated after a certain period (i.e.
every 24 hours), or after a maximum allowable head loss is achieved
(i.e. 1 m), or when the effluent solids reach a threshold level
(i.e. 10 g/m3). This last mode may be useful
for optimizing plant performance in the mathematical model. In
manual mode, the operation of the BAF has to be done through a
Control window (or file input), which needs to be set up with all
the operational variables.
Further model
parameters are included in the mass transport,
stoichiometric, kinetic and filtrationmenu items in the
BAF model. These forms are described in the Trickling
Filter Model section of this chapter and in the Sand
Filtration Models section developed in CHAPTER
9.
Figure 7‑19 - More Advanced BAF
Operational Parameters
Output Variables –
Hydraulics
Here, hydraulic
information can be accessed and displayed for each of the
horizontal filter sections (6 by default). It includes dilution
rate and headloss-related variables: Reynolds number, friction
factor, and headloss in each horizontal filter section for both
clean and dirty filter beds.
2D BAF
The solids capture
rate during filtration can be displayed for each of the horizontal
filter sections. The other variables (O2 partial
pressure, pore size, biofilm thickness, deposit concentration, bulk
deposit volume) can be displayed on 3D graphs. The x-axis has the
six horizontal filter sections, the y-axis has the five biofilm
layers and liquid phase, and the z-axis displays the variable
value.
Filter Segment
The state variables
and the solids composite variable can be displayed on 3D graphs.
The x-axis has the six horizontal filter sections, the y-axis has
the five biofilm layers and liquid phase, and the z-axis displays
the variable value.
Liquid Film Concentration
in Filter Part
In this form, the
state variables and unattached solids composite variable values in
the liquid phase can be selected for display for each of the
horizontal filter sections. The values displayed are only valid for
the BAF units in filtration mode.
Variables in Backwashed
Segment
In this form, the
state variables can be selected for display for each of the biofilm
layers. The values displayed are only valid for the units in
backwash mode, and represent the average of all horizontal filter
sections. In filtration mode, the filter is treated as a completely
mixed tank.
The Hybrid-System
model in GPS-X is based on a combination of the standard plug flow
tank configuration with suspended growth biomass, and the GPS-X
biofilm model representing fixed film growth on the media inserted
into the tank. Any type of media will be represented in the model
as long as the specific surface area is set correctly. Thus the
model is able to represent commercial systems such as MBBR,
Ringlace, Captor, Bionet and other types of hybrid systems, whether
they contain sludge recycle or not.
Biological Component
The hybrid-system
uses the same biological reactions found in the suspended-growth
models. For more specific information on biological models, please
consult CHAPTER 6.
Biofilm Component
The existing
biofilm model in GPS-X is based on Spengel and Dzombak
(1992). This model was adapted to the hybrid-system. The
model handles soluble material diffusion, biofilm growth, and
particulate attachment and detachment. Details are described in the
Trickling Filter Model section of this chapter.
Model Parameters
This section of the
chapter discusses the various model parameters and inputs that the
user will encounter when using this model.
Physical
The physical parameters are found under the Parameters
sub-menu item Physical. It contains physical dimensions to
describe the hybrid-system, and model dimensions which allow the
user to specify how the physical system will be modelled.
The hybrid-system
requires inputs such as the specific surface of media, the
water displaced by media, and the specific density of
media.
Operational
In the Operational form, aeration constants are
available similar to all aerated biological units. Specific to the
hybrid-system is the ability to specify two additional recycle
streams. These streams are internal recycle with carrier(if
internal recycle carries carrier with it) and flow from tank #
with carrier (if the carrier can flow from one cell to the next
one).
Further model parameters are
included in the mass transport, stoichiometric and
kinetic menu items in the hybrid-system model. (See
Trickling Filter Model section in this chapter)
Initialization
In addition to the
standard initial volume input, the hybrid-system requires the
reactor portion filled by media (media or empty bed fill),
which can be input in the Initial Volume sub-menu form.
Output Variables
In addition to the
standard effluent parameters, there are a number of fixed film
specific variables that can be displayed. These are accessed
through the object's output variables menu item.
The denitrification
filter objects are found in the “Tertiary Treatment” group of
objects, rather than in the “Attached Growth” group, as the other
biofilm objects.
The denitrification
filter model is similar to the trickling filter model in hydraulic
structure. There are, however, a few fundamental
differences:
1.
The filter is assumed to be entirely flooded. There is no
empty void space between media, as in the trickling filter
model.
2.
The trickling filter model is assumed to be downflow, whereas there
are two different denitrification filter models: Upflow and
Downflow, as shown in Figure 7‑20.
Figure 7‑20
- Upflow and Downflow Denitrification Filter Objects
Biological Component
The denitrification
filter model uses the same biological reactions found in the
suspended-growth models. For more specific information on
biological models, please consult CHAPTER 6.
Biofilm Component
The existing
biofilm model in GPS-X is based on Spengel and Dzombak
(1992). This model was adapted for use in the
attached-growth objects such as the trickling filter and
denitrification filters. The model handles soluble material
diffusion, biofilm growth, and particulate attachment and
detachment. Details are described in the Trickling Filter
Model section of this chapter.
Filtration Component
The filtration of
solids is modeled as a simple capture of the particulate state
variables from the effluent layer of the denitrification filter
(bottom layer in the downflow filter and top layer in the upflow
filter). Users specify a capture fraction, and all solids
captured are removed through the backwash connection stream.
Backwashing (and its associated removal of solids) is modeled as a
continuous process flow.
Model Parameters
This section of the
chapter discusses the various model parameters and inputs that the
user will encounter when using this model.
Physical
The physical parameters are found under the Parameters
sub-menu item Physical. It contains physical dimensions to
describe the denitrification filter, and model dimensions which
allow the user to specify how the physical system will be
modelled.
The denitrification
filter requires inputs such as the specific surface of media
and the porosity of media.
Operational
The operational
menu contains three parameters that define the removal of solids
via backwash. The solids capture fraction defines the
fraction of particulate state variables (on a concentration basis)
that is removed from the effluent stream. The
parameters backwash duration during 24-hr period and
backwash flow are used to determine the total amount of flow
leaving through the backwash connection point per day. The
backwash concentration is calculated as the total amount of
captured solids divided by the daily backwash flow rate.
Other Menus
Further model
parameters are included in the mass transport,
stoichiometric and kinetic menu items in the
denitrification filter model. These forms are described in the
Trickling Filter Model section of this chapter.
Initialization
The initialization
menu of the denitrification filter object is the same as for the
trickling filter object, and contains the initial concentrations of
the state variables in each layer of the filter.
Output Variables
In addition to the
standard effluent parameters, there are a number of fixed film
specific variables that can be displayed. These are accessed
through the object's output variables menu item.
The
Membrane-Aerated (MABR) object is found in the Attached Growth
section of the Unit Process Table.
Figure 7‑21- Hollow Fibre Membrane
Aerated Bioreactor object
The MABR object
uses a structure similar to the Hybrid object, in that it is a
suspended growth reactor with media supporting biofilm
growth. Unlike the Hybrid object, the media can only be fixed
(as opposed to floating), and the aeration of the biofilm is done
from the membrane surface itself (as opposed to via the bulk
liquid), effectively aerating the biofilm from the inside
out. Regular diffused aeration of the bulk liquid is also
available in this object.
Figure
7‑22 shows the structure of the MABR biofilm model.
At the left, the media surface area supports the growth of
biofilm. The biofilm is modelled as 5 homogeneous layers,
followed by an interface with the bulk liquid. Soluble
components from the bulk liquid can diffuse into (and out of) the
outermost biofilm layer (the one in contact with the liquid).
The MABR membrane surface allows for the diffusion of oxygen
through the membrane and into the innermost biofilm layer.
Figure 7‑22 -
MABR biofilm structure, showing diffusion of soluble components
As the with the
other biofilm models in GPS-X, each of the biofilm layers is
modelled as a completely-mixed reactor. All aspects of solids
transfer (internal solids exchange, attachment and detachment at
the liquid/biofilm interface) are all modelled similar to the
hybrid model.
Model Parameters
This section of the
chapter discusses the various model parameters and inputs that the
user will encounter when using this model.
Physical
The physical parameters are found under the Parameters
sub-menu item Physical. It contains physical dimensions of
the MABR unit, and details of the media surface area and biofilm
properties.
The total amount of surface area available for growing biofilm is
equal to the number of cassettes multiplied by the
modules per cassette, cords per module and media
length. The surface area per length of media is
determined from the media outside diameter and the thickness
of the biofilm. Therefore, the surface area per length of
media increases with increasing media diameter and biofilm
thickness.
Figure 7‑23 - MABR Physical Parameters
Menu
Operational
The operational
menu contains parameters that define the operation of the membrane
aeration, conventional bulk liquid aeration and hydraulic settings
(e.g. internal mixed-liquor recycle flow) of the MABR unit.
The first section
of the menu, “Inner Biofilm Aeration” contains settings for
specifying the mass transfer of oxygen from the membrane to the
innermost biofilm layer. There are three options available
for specifying oxygen transfer:
1)
setting the mass transfer coefficient, kLa, in units of
1/d
2)
setting the mass transfer of oxygen directly, in kg/d
3)
using the pressure difference equation, which specifies the oxygen
transfer through the length of the membrane using inlet and outlet
pressure and oxygen fraction
The
pressure-difference oxygen transfer model is described by Côté
(1989), and is shown below:
where,
J =
oxygen flux, mol/m2-sec
K = mass transfer
coefficient, m2/sec
pin = partial pressure of
oxygen at inlet, Pa
pout = partial pressure of oxygen at
outlet, Pa
H = Henry’s Law
constant, Pa-m3/mol
CL = oxygen
concentration in liquid, mol/m3
The operational
menu for selecting the method for specifying oxygen transfer is
shown below.
Figure 7‑24 - MABR Operational Menu
Note that the
kLa and pressure difference options incorporate the
oxygen saturation term into the overall mass transfer, meaning that
as the dissolved oxygen in the liquid (or biofilm, in this case)
comes closer to saturation, the oxygen transfer decreases.
This is not the case for the direct oxygen mass transfer
setting. The amount of oxygen (in mass-per-unit-time terms)
will be transferred, regardless of oxygen saturation. The
user is should note that this can possibly result in the oxygen
concentration in the innermost biofilm layers being above
saturation.
The default setting
for oxygen transfer specification is the pressure difference
model.
Mass Transport Menu
The mass transport
menu of the MABR object contains parameters relevant to the
transport of components to, from, and within the layers of the
biofilm. The menu is shown below:
Figure 7‑25 - MABR Operational Menu
These parameters
are similar to those used in other biofilm objects:
attachment
rate – a rate constant used to determine the mass flow of
solids from the bulk liquid to the outermost biofilm layer
detachment
rate – a rate constant used to determine the mass flow of
solids from the outermost biofilm layer to the bulk liquid
anoxic shear
reduction factor – a parameter which reduces the amount of
detachment in tanks with no aeration, to simulate the calmer
conditions present without diffused aeration
internal solids
exchange rate – a parameter used to determine the mass flow of
solids between internal layers in the biofilm
For details on how
the attachment/detachment and solids exchange rates are calculated,
please see the details of the trickling filter model.
The Mass Transport
menu also contains further constants for the diffusion of soluble
components through the biofilm. These are accessed by
clicking on the MORE… button at the bottom of the
menu. The menu that pops up contains diffusion
constants for each of the soluble components as well as the
biofilm reduction factor. The reduction in diffusion
constant for each soluble component is treated differently in
the MABR model than it is in the other biofilm models in
GPS-X. In the other models, there is one biofilm reduction
factor that is applied to all the soluble components (default =
0.5). In the MABR model, each soluble component has its own
biofilm reduction factor, which can be different from each
other. In general, the default values for the different
soluble components are 0.2 for organic components, and 0.8 for
inorganic components, as per Stewart (2003).
Other Menus
Further model
parameters are included in the stoichiometric and
kinetic menu items. These forms are described in the
Trickling Filter Model section of this chapter.
Initialization
The initialization
menu of the MABR object is the same as for other biofilm objects,
and contains the initial concentrations of the state variables in
each layer of the biofilm.
Output Variables
In addition to the
standard effluent parameters, there are a number of fixed film
specific variables that can be displayed. These are accessed
through the object's Output Variables menu item.
The MABR
Performance Variables menu contains several display unique state
variables that illustrate the mass transfer and concentrations of
various elements throughout the biofilm. The menu is shown
below.
Figure 7‑26 - MABR Performance Variables
Menu
The 1-D
Rates section contains output variables such as nitrification
rate and oxygen transfer rate, calculated in various ways.
The Ammonia Fate section summarizes the transformation of
ammonia in the reactors of the MABR unit (e.g. how much is removed
biologically, how much diffuses into the biofilm, and how much
remains in the bulk liquid).
The 1-D
Masses and 1-D Mass Flows summarize the distribution and
fate of the heterotrophic, ammonia-oxidizing, and nitrite-oxidizing
biomass in the bulk liquid and the biofilm. The 2-D
Concentrations section contains 2-dimensional array variables
of the various soluble and particulate components in the
biofilm. These variables are intended to be displayed on a
“3D Bar Graph” type, and appear as shown below:
Figure 7‑27 - MABR Performance Variables
Menu
In these diagrams,
the orientation of the media, bulk liquid and the biofilm layers is
shown below:
Figure 7‑28 - MABR Performance Variables
Menu
CHAPTER 8
This chapter
provides a description of the sedimentation and flotation models
available in GPS-X.
Sedimentation is one of the most important unit processes in
activated sludge treatment plants. The sedimentation unit, whether
it is a primary settler, secondary clarifier, or sludge thickener,
provides two functions: clarification and thickening. The primary
settlers and sludge thickeners are designed and operated to take
advantage of the thickening process, while the secondary clarifiers
are designed and operated to take advantage of the clarification
process.
In GPS-X, the sedimentation models
are either zero- (point) or one-dimensional (1d suffix), and either
reactive (mantis, asm1...) or nonreactive (simple).
The following models are available:
·
Zero-dimensional, nonreactive: point
·
One-dimensional, nonreactive: simple1d
·
One-dimensional, reactive: mantis, asm1,
asm2d,asm3, newgeneral
In reactive models,
biological reactions are included, and the model names are
associated with the corresponding suspended-growth models,
described in Chapter 6 (page 95). For example, the
mantis sedimentation model uses the mantis
suspended-growth model.
In the
one-dimensional models, the settler is divided into a number of
layers (10 by default) of equal thickness, as depicted in
Figure 8‑1.
The following
assumptions are made:
1.
Incoming solids are distributed instantaneously and uniformly
across the entire cross-sectional area of the feed layer.
2.
Only vertical flow is considered.
Figure 8‑1 -
One-Dimensional Sedimentation Model
The models are based on the solids flux concept: a mass balance is
performed around each layer, providing for the simulation of the
solids profile throughout the settling column under both
steady-state and dynamic conditions.
Table 8‑1
shows the appropriate contribution of each layer of the settler to
the mass balance. There are five different groups of layers,
depending on their position relative to the feed point. This is
shown schematically in Figure 8‑2.
The models are based on traditional solids flux analysis, but the
solids flux in a particular layer is limited by what can be handled
by the adjacent layer.
The solids flux due to
bulk movement of the liquid is a straightforward calculation based
on the solids concentration times the liquid bulk velocity, which
is up or down depending on its position relative to the feed
layer.
Table 8‑1 –
Sedimentation Model: Input-Output Summary
|
Input
|
Output
|
Layer
|
Feed
|
Settling
|
Bulk Liquid Flux
|
Settling
|
Bulk Liquid
Flux
|
Top
|
-
|
-
|
up
|
+
|
up
|
Layers above feed point
|
-
|
+
|
up
|
+
|
up
|
Feed
|
+
|
+
|
-
|
+
|
up-down
|
Layers below feed point
|
-
|
+
|
down
|
+
|
down
|
Bottom
|
-
|
+
|
down
|
-
|
down
|
Note: + = phenomenon considered; - = phenomenon not
considered
|
Figure 8‑2 -
Solids Balance Around the Settler Layers
The solids flux due to bulk movement of the liquid is a
straightforward calculation based on the solids concentration times
the liquid bulk velocity, which is up or down depending on its
position relative to the feed layer.
The solids flux due to
sedimentation is specified by a double exponential settling
function, applicable to both hindered sedimentation and flocculant
sedimentation conditions. The settling function, described by
Takács et al. (1991), is given by:
Equation 8.1
where:
vsj
= the settling velocity in layer j (m/d)
vmax = the maximum
Vesilind settling velocity (m/d)
rhin = hindered zone settling
parameter (m3/gTSS)
rfloc = flocculant zone settling
parameter (m3/gTSS)
X
jo =
Xj – Xmin, where
Xmin is the minimum attainable suspended solids
concentration, Xj is the suspended solids
concentration in layer j
The minimum attainable solids concentration in a layer,
Xmin, is calculated as a fraction
(non-settleable fraction or fns) of the influent
solids concentration to the settler:
Equation 8.2
It is subject to a maximum value specified by the user; the
maximum non-settleable solidsor Xminmax.
The settling velocity is lower bounded to zero, so that if the user
specifies parameter values that would result in settling velocities
becoming negative, a warning message is printed in the simulation
Log window. The settling velocity is also subject to a
maximum value specified by the user; the maximum settling velocityor
vbnd.
The settling function is shown in Figure 8‑3. The
four regions depicted in this figure are explained as follows: I)
the settling velocity equals zero, as the solids attain the minimum
attainable concentration; II) the settling velocity is dominated by
the flocculating nature of the particles; thus the settling
velocity is sensitive to the rfloc parameter; III) settling
velocity has become independent of solids concentration (particles
have reached their maximum size; and IV) settling velocity is
affected by hindering and becomes dependent on the rhin
parameter (the model reduces to the Vesilind equation).
Figure 8‑3 –
Settling Velocity vs. Concentration
Simple Models
In the
simple1d sedimentation model, the only numerically
integrated variable is the suspended solids concentration. This
model can be used when biological reactions in the settler can be
ignored. The concentrations of particulate state variables in the
influent to the settler (heterotrophic organisms, etc.) are stored
as fractions of the total suspended solids concentration entering
the settler. Once the model completes the numerical integration of
the suspended solids in the settler layers (at the end of each
numerical integration time step), the concentrations of particulate
state variables in the effluent, underflow (RAS), and pumped flow
(WAS) are restored using those fractions. The concentrations of
soluble state variables are not changed in the simple1d
model.
Sludge Blanket Threshold
Concentration
This variable is
used to define the sludge blanket height for display purposes. If
the concentration in a settler layer is above this threshold value
(searching from top to bottom layer), then the sludge blanket is
defined as the height of that layer.
Critical Sludge Blanket
Level
This parameter is
used to define the height of the sludge blanket in order for the
dissolved oxygen in the underflow and pumped streams to be
zero.
Soluble and Particulate
Components
The soluble state
variables in the nonreactive models are subject to a complete mix
zone, unlike the particulate components, which move from layer to
layer. If the user wishes to subject the soluble components to a
number of tanks in series, then they must select a reactive type
settler. In this case all of the state variables are transported
from cell to cell according to the bulk fluid motion (but only the
particulate components will be affected by a settling term). The
feed layer then will become an important term in fixing the number
of layers in series through which the soluble components will
flow.
Correlation to Sludge
Volume Index and Clarification
A feature of the
sedimentation models in GPS-X (secondary clarifiers only) is the
correlation provided between the settling parameters and Sludge
Volume Index (SVI) measurements.
The SVI test characterizes the
sedimentation, which occurs in the high solids concentration band
of a clarifier. To specify the sedimentation characteristics over
the full concentration spectrum, another parameter, the
Clarification factor, is needed to specify the settling
behavior in the flocculant or low solids concentration regions.
This factor is a relative clarification index; a high number (1.0)
indicates good clarification and a low number (0.1) indicates poor
clarification.
The correlation
equations are:
Equation 8.3
Equation 8.4
Equation 8.5
where:
vmax
= maximum Veslind settling velocity(m/d)
rhin
= hindered zone settling parameter (m3/gTSS)
rfloc
= flocculant zone settling parameter (m3/gTSS)
SVI
= sludge volume index (mL/g)
clarify
= clarification factor
fcorr1 -
fcorr9 = SVI correlation coefficients
To use the
correlation, the user must set the parameter use SVI to estimate
settling parametersto ON. The Sludge Volume Index
and Clarification parameters can then be specified and the
settling parameters are calculated automatically by GPS-X. The
correlation coefficients can be accessed in the
Options > General Data > System > Parameters >
Miscellaneous form.
Note:
The
default values of the correlation factors are for SVI. A
correlation with SSVI has not been performed. The default values of
the correlation coefficients are based on five sewage treatment
plants.
Influent Flow
Distribution
The sedimentation
models will account for hydraulic effects caused by an increase in
influent flow. Flow conditions are considered normal when the
influent flow divided by the surface area is less than the
quiescent zone maximum upflow velocity specified by the
user. The load to the settler under normal flow conditions enters
the settler at the feed layer. However, as the influent flow
increases, the load to the settler is distributed to the layers
below the feed point. When the upflow velocity in the settler
surpasses the complete mix maximum upflow velocity specified
by the user, the entire load enters the bottom of the settler.
When the upflow velocity is
between the quiescent
zone maximum upflow velocity and complete mix maximum upflow velocity, a
smooth transition of feed distribution between the low loading case
and the high loading case is generated. An average (medium) loading
condition is initially calculated where the upflow velocity
(vuavg) is the average of the quiescent zone maximum upflow
velocity(vumin) and the complete mix maximum upflow
velocity(vumax). At this hydraulic loading, the input
distribution is equal to the feed layer and all layers below.
Smooth distribution is achieved by
two linear interpolations. The first interpolation is between
vumin and vuavg. At vumin, the influent
fraction to the feed layer is 1.0 (all flow enters the feed layer),
while at vuavg the influent fraction to the feed layer is
1.0 divided by the number of layers below the feed layer (including
the feed layer itself). Once the influent fraction to the feed
layer is calculated, the remaining flow is equally distributed to
the layers below.
A similar algorithm will ensure
smooth flow distribution above vuavg. The algorithm first
calculates the feed fraction to the bottom layer. If vu is higher
than vuavg, then the algorithm distributes the rest
evenly.
This procedure of flow
distribution is modelling the feed distribution to the settler by
the influent baffle. The feed distribution is shown in Figure
8‑4.
During higher
flows, the momentum of the incoming flow tends to carry the load
further past the bottom edge of the influent baffle, effectively
changing the feed point in the settler. The flow distribution
aspect of the model captures this phenomenon.
Figure 8‑4 –
Load Distribution into Settler
There are four
types of settler objects in GPS-X:
1.
rectangular primary
2.
circular primary
3.
rectangular secondary
4.
circular secondary
All of the settler
objects contain the same settling model. The difference
between primary and secondary clarifiers is the default settling
parameters and the SVI correlation (SVI correlation is not used in
the primary settlers).
The difference
between the circular and rectangular configurations is restricted
to the specification of the area and/or geometry of the
tanks. The rectangular tanks require only a total surface
area for input, whereas the circular tanks have several shapes
available.
Circular Settler
Configurations
There are 4
different circular settler shapes available in the circular primary
and secondary settler objects, as shown in Figure
8‑5. The surface area of each layer is calculated from the
clarifier type and dimensions specified by the user. For example,
as shown in the figure, a clarifier with a conical shape will have
a larger surface area in the upper layer and a smaller surface area
in the bottom layer.
Figure 8‑5 –
Circular Settler Shapes
This section of the
chapter describes the flotation model (simple1d) associated
with the Dissolved Air Flotation (DAF) unit.
The flotation model is closely
related to the one-dimensional sedimentation model. It is partly
based on the double-exponential function, but the model is inverted
to promote flotation of solids as opposed to sedimentation of
solids. The flotation model includes a solids flux component to
account for floating of solids. This floating component is
primarily controlled by an air-to-solids ratio and a polymer
dosage, specified by the user.
One-Dimensional Model
The one-dimensional
flotation model will predict the amount of solids removal achieved
by the DAF unit. Solids are removed from the top of the unit in the
float stream. Effluent is removed from the bottom of the DAF unit
in the effluent stream. The one-dimensional flotation model is
primarily based on the solids flux theory presented previously in
this chapter, but it is modified to account for flotation as
opposed to sedimentation. The major difference between the
flotation and sedimentation models is the direction of solids flux
and the parameters controlling it.
In the
one-dimensional flotation model, the DAF unit is divided into a
number of layers (10 by default) of equal thickness, depicted in
Figure 8‑6.
The following
assumptions are made:
1.
Incoming solids are distributed instantaneously and uniformly
across the entire cross-sectional area of the feed layer.
2.
Only vertical flow is considered.
Figure 8‑6 -
Layered Flotation Model
The model is based
on the solids flux concept: a mass balance is performed around each
layer, providing for the simulation of the solids profile
throughout the DAF unit under both steady-state and dynamic
conditions.
The model is based on traditional
solids flux analysis, with an additional component for flotation.
Model detail is provided in the previous section of this chapter
devoted to one-dimensional sedimentation models.
The solids flux due to the bulk
movement of the liquid is calculated by multiplying the solids
concentration by the liquid bulk velocity (flow divided by area),
which may be up or down depending on the position relative to the
feed layer.
The solids flux due
to flotation is specified by the same double exponential function
used for sedimentation (Equation 8.1). The parameters
within Equation 8.1 are altered to account for
effects of flotation:
vsj
= floating velocity in layer j (m/d)
vmax = floating
velocity with optimal air-to-solids ration (m/d)
rhin = hindered zone floating
parameter (m3/gTSS)
rflo = free floating
zone floating parameter (m3/gTSS)
X
jo =
Xj – Xmin, where
Xmin is the minimum attainable suspended solids
concentration, Xj is the suspended solids
concentration in layer j
The minimum attainable solids concentration in a layer,
Xmin, is calculated as a fraction
(non-floatable fraction of fns) of the influent
solids concentration to the DAF unit:
Equation 8.6
It is subject to a
maximum value specified by the user, the maximum non-floatable
solids or Xminmax. The floating velocity is
lower bounded to zero, so that if the user specifies parameter
values that would result in floating velocities becoming negative,
a warning message is printed in the simulation Log window.
The floating velocity is also subject to a maximum value specified
by the user, the maximum floating velocityor
vbnd.
Model Parameters
This section of the
chapter contains a description of the various model parameters and
inputs that the user encounters when using the flotation model.
Physical
These menu items
are found under the Parameters sub-menu item
Physical. This input form contains the actual physical
dimensions of the DAF unit being modelled, including the tank
surface area, the maximum water level or height of
the tank, and the location of the feed point relative to the bottom
of the tank. The fourth item on the form allows the user to define
the number of equivalent layers contained within the model. The
physical parameter form is shown in Figure 8‑7.
Figure 8‑7 –
Physical Parameters for the DAF Unit
Operational
The operational
parameters for the DAF unit are located within the
Parameters sub-menu item Operational. This input form
contains the polymer dosage (g polymer/kg solids) and the
air-to-solids ratio (g air/g solids) for daily operation of
the DAF unit. The maximum float flow is an upper boundary
limit for the amount of float that can be removed from the DAF
unit. The operational input form is shown in Figure
8‑8.
Under the
sub-heading other parameters, the optimal polymer
dosage and optimal air-to-solids ratio are defined.
These two values provide an estimate of the best expected solids
condition. The optimal values may be expressed based on past
operating experience with the actual DAF unit or may be provided
through manufacturer literature. The dry material content of the
float at the optimal polymer dosage is defined as a percentage. The
dry material content of the float without polymer dosage is also
defined. The model will use the range between these two values to
predict the dry material content of the float under actual
operating conditions.
Figure 8‑8 -
Operational Parameters for the DAF Unit
Floatation
The
Flotation parameters, shown in Figure 8‑9, are
used to define the variables contained within the model (Equation
8.1). The maximum floating velocity provides an upper
boundary limit for the model. The next variable is the expected
floating velocity under the optimal air-to-solids ratio, as
defined under the operational input form. The hindered zone
and free floating zone floating parameters control the
floating velocity defined in Equation 8.1. There will
likely be a fraction of solids, which cannot be removed by
flotation. This fraction is characterized as the non‑floatable
fractionand an upper boundary for this parameter as a
concentration is set as the maximum non-floatable
solids.
Figure 8‑9 –
Flotation Parameters for the DAF Unit
CHAPTER 9
This chapter
describes the deep bed granular filtration models available in
GPS-X. The continuous and massbalance models are
based on empirical removal efficiencies. The simple1d model
is a mechanistic model based on continuity and kinetic equations
that describe the removal of suspended particles by deep bed
granular filters.
The basis for the
continuous model is the direct specification of the filter
performance through two parameters: the backwash flow
fraction and the backwash solids mass fraction
(Figure 9‑1). The backwashout
connection (bottom left of object) is not used in this
model.
The backwash flow
fraction (frqbw) is the fraction of the incoming flow to the
filter (input connection - top left of object) that is used for
backwash. Using this parameter, GPS-X will calculate a
continuous backwash flow (Qb) associated
with the backwashout stream (top right of the object):
Equation 9.1
The continuous output flow (bottom right of the object), is then
calculated from the difference between the input flow and the
backwash flow:
Equation 9.2
The backwash
solids mass fraction (frxbw) is the fraction of incoming solids
that is captured by the filter, ending up in the backwash
stream. Using this parameter, GPS-X calculates the solids
concentration in the backwash stream
(Xb):
Equation 9.3
The solids
concentration in the output stream (Xo) is
calculated from:
Equation 9.4
Figure 9‑1 –
Operational Parameters Form – Continuous Model
The massbalance model is based on empirical removal
efficiency for suspended solids, BOD and TKN parameters. A `best'
value is set for removal efficiency immediately after backwash. The
removal efficiency decreases during the filter cycle time as the
filter is fouled until the next backwash. The decrease in the
removal efficiency is modelled with an exponential function:
Equation 9.5
where:
out
= filter effluent component concentration (mg/L)
in
= filter influent component concentration (mg/L)
bestefficiency = best removal efficiency after
backwash (%)
foulingcoeff = fouling
coefficient
cycletime
= time elapsed since last backwash (h)
A minimum efficiency value for suspended solids removal is also
defined as an operational parameter. This parameter does not affect
removal computed by the model but warns the user that, once removal
of solids gets below this minimum value, the filter is undergoing
high headloss.
Figure 9‑2 –
Operational Parameters Form – Massbalance Model
The massbalance model accounts for the mass accumulated in
the filter during the filter cycle time. The accumulated mass is
removed during backwash. Therefore, stoichiometric parameters for
the filter effluent and the backwash may be specified. The
operational parameters for the massbalance model are shown
in Figure 9‑2.
The basis for the simple1d model is the combination of the
continuity (mass balance) and the kinetic partial differential
equations by Horner et al. (1986), which describe the removal of
suspended particles by a granular filter:
Equation 9.6
where:
s
= volume of deposited solids per unit bed volume
l
= filtration coefficient
C
= concentration of suspended particles at depth L and time
t
u
= approach velocity (velocity of the fluid above the filter
bed)
ed
=
porosity of deposited solids
t
= time
When combined with defining equations for the deposited (attached)
solids (X= s×d) and the unattached solids
(X = C ×dd) in the filter, the following
equation is derived for the simple 1d model:
Equation 9.7
where:
X
= unattached solids
Xd
= attached (deposited)
solids
δd =
density
The filter bed is
divided into layers and it is assumed that the specific deposit is
uniform across each layer. During the backwash cycle, the
average deposit through all layers of the filter is used. The
average deposit after backwash is used as the initial condition for
the subsequent filter run.
Model Parameters
This section
discusses the various model parameters and inputs that the user
encounters when using the simple1d model.
Physical
These menu items
are found under the Parameters sub-menu item Physical
and contain both real physical dimensions to describe the actual
filter modelled, and model dimensions which allow the user to
specify the number of layers and an effective particle diameter.
The real dimensions of the filter are the bed surface area and the
bed depth.
Operational
Operational items,
found under the Parameters sub-menu, include the basis for
the duration of the filtration run. The duration of the filtration
run may be based on a user-specified time, headloss or effluent
suspended solids concentration. The filtration run may also be set
manually. Other operational parameters include the influent flow
and specification of the backwash parameters (duration, rate).
Stoichiometric
The stoichiometric
fractions available for the model are the ratios of particulate COD
to VSS, VSS to TSS and BOD5 to
BODultimate.
Filter
The filtration
constants required for the model are shown in Figure
9‑3. The filtration constants include the clean bed
filtration coefficient (lo), the initial
porosity of the filter bed (eo) and the porosity
of deposited solids (ed), the ultimate
bulk specific deposit (su), and the density
and dry material content of the solids. The packing factor is
used in the defining equation for the variation of the filtration
coefficient with the bulk specific deposit as given by Ojha and
Graham (1992).
Figure 9‑3 -
Filter Parameters
Output Variables
The flow and
characteristics of the filter flow and backwash can be displayed on
output graphs. There are various filter variables available in the
sub-menus under Output Variables. The output variables
include dilution, headloss and rates, the
unattached and attached solids in the filter and
filter conditions (i.e. bulk deposit, bed porosity).
CHAPTER
10
This chapter describes the aerobic and
anaerobic digestion models used in GPS-X.
This section
describes the basic anaerobic digestion model associated with the
anaerobic digester object. It is a modified version of the model
developed by Andrews (1969), and
Andrews et al. (1971). The
modifications to the original model are:
·
The addition of temperature sensitivity for the hydrolysis of
volatile suspended solids (VSS) and the growth of methanogenic
organisms. The Arrhenius equation is used with a base temperature
of 35 degrees Celsius.
·
The chemical equilibria were modified by introducing Hydromantis'
pH model (See PH Tool section in CHAPTER
12).
·
The introduction of particulate inert inorganic material (xii).
This component remains unchanged within the digester and is
introduced for the sole purpose of assessing its impact on other
processes downstream of the digester.
·
The addition of a rate for toxic substance degradation.
Conceptual Model
The basic
anaerobic digestion model consists of two reactors: one for the
liquid phase and the other for the gaseous phase. Both are modelled
as completely mixed reactors. Transfer of gaseous products between
the liquid phase and the gaseous phase is modelled using a standard
two-film mass transfer equation. Gaseous carbon dioxide
(CO2) and methane (CH4) are
assumed to follow the ideal gas law. No further reactions take
place in the gaseous phase, which has a total gas pressure of
760 mm of Hg (i.e., atmospheric pressure).
Figure 10‑1
- Schematic Diagram of the Anaerobic Digestion Model
A schematic diagram
of the anaerobic digestion process is shown in Figure
10‑1, where:
In the gas phase:
qtg: = total gas flow (m
3/d)
qco2 = CO2 gas flow (m
3/d)
qch4: = CH4 gas flow (m
3/d)
gco2: =
partial pressure of CO2 (atm)
gch4 = partial pressure of CH4
(atm)
In the liquid phase:
State
Variables
xmh: = methanogens
(gCOD/m3)
vss =
volatile suspended solids (gCOD/m 3)
slf = total
volatile fatty acids (gCOD/m3)
sco2t = total soluble
CO2(moles/L)
sz = net
cations (moles/L)
stox = toxic substance
(g/m3)
Composite
Variables
snhn = free ammonia (moles/L)
snhi = ionized ammonium
(moles/L)
slfn = non-ionized
volatile fatty acids (moles/L)
slfi
= ionized volatile fatty acids (moles/L)
pH = pH
hco3 = bicarbonate (moles/L)
h2co3 = carbonic acid (moles/L)
co2 =
carbonate (moles/L)
salk = alkalinity
(g/m3)
Mathematical Model
The mathematical
equations used in the liquid and gaseous phases are provided in the
corresponding Model matrix form in Appendix A. The general reaction
pathway is shown in Figure 10‑2.
Figure 10‑2-
General Reaction Pathway
Stoichiometry
The relative
amounts of chemical components produced by the biological and
chemical reactions in the anaerobic digester are specified by
stoichiometric coefficients. The chemical reaction stoichiometry is
defined by the balanced chemical reaction equations. The
biological stoichiometry is defined by the yield coefficients.
Seven yield
coefficients are used by the basic model:
ya = slf / vss
yb =
sco2t / vss
yc = xmh /
slf
yd =
sco2t / xmh
ye = gch4 /
xmh
yf
= snhi / vss
yg
= xmh / snhi
In each case the yield is defined as the ratio of the change in a
product to the change in a reactant. Volatile acids and volatile
suspended solids are expressed as their equivalent chemical oxygen
demand (COD).
Components and Processes
in the Mathematical Model
Volatile
Suspended Solids
The rate of
hydrolysis of VSS is assumed to be first order with respect to the
concentration of VSS. A temperature correction factor
(ftkco) is calculated using the Arrhenius equation with a
base temperature of 35 degrees Celsius. The temperature
correction factor is incorporated in the equation for the
hydrolysis rate:
Equation 10.1
where:
kco = rate constant for
the hydrolysis of VSS
Methanogenic
Organisms
The growth of the
microorganisms responsible for the generation of methane is
modelled using the Monod equation modified by switching functions
(similar to the IWA models). The rate of growth of methane
producing bacteria (r2) is assumed to be proportional to
their concentration (mumh). The model uses un‑ionized
volatile acids (slfn) as the substrate and incorporates two
switching functions for inhibition: one for inhibition by
slfn and the other by free ammonia (snh).
The resulting rate
equation is:
Equation 10.2
where:
ftmum
= the temperature correction factor for the growth of
methanogens
mumh
= maximum specific growth rate for methanogens
ks
= the half-saturation coefficient
kia,
kin
= the inhibition constants for slfn and snh,
respectively
As in other
biological models, the rate of decay of methane producing bacteria
(r3) is assumed to be proportional to their
concentration:
Equation 10.3
where:
kd = the decay
rate coefficient
The effect of toxic substances (stox, input as a special
component to the influent parameters) is taken into account by
using a first order expression for the rate of inactivation of
methane bacteria:
Equation 10.4
where:
ktox = the inactivation rate
coefficient
Based on the above rates, the net rate of generation of methanogens
(rxmh) is:
Equation 10.5
Toxic
Substances
A rate of toxic substance degradation (r5) is incorporated
in the basic model. The rate of degradation is assumed
to be first order with respect to the concentration of toxic
substances:
Equation 10.6
where:
kb = toxic
substance degradation rate
Total Volatile
Fatty Acids
The kinetic expression for total volatile fatty acids (slf)
can be established using the previously presented kinetic
expressions and appropriate yield coefficients:
Equation 10.7
where:
(ya
r1)
= rate of generation of slf by hydrolysis
‑r2/yc
= rate of utilization of slf by the growth of
methanogens
Methane
The biological
generation of methane can be expressed in terms of bacterial growth
rates and yield coefficients. The model assumes that the solubility
of methane is negligible and all methane generated is immediately
transferred to the gas phase:
Equation 10.8
where:
vm = volume of the
liquid phase in the digester
Carbon
Dioxide
The mass transfer of carbon dioxide between the liquid and gas
phases is calculated using the standard two-film gas transfer
equation.
Equation 10.9
where:
klac2o2
= mass transfer coefficient for CO2
co2sat
= saturation concentration of CO2 in the liquid phase,
and H2co3=sco2t-hco3-co2
The pH model within the basic digester model calculates
bicarbonate (hco3) and carbonate (co2). The
concentration of dissolved CO2 in the liquid
phase at equilibrium (co2sat) is calculated using
Henry's law:
Equation 10.10
where:
henryco2
= Henry’s law constant for CO2
gco2
= partial pressure of CO2 in the gas phase
Combining the above equations:
Equation 10.11
The rate of biological generation of total soluble carbon dioxide
(rsco2t) can be expressed in terms of bacterial
growth rates and yield coefficients. Combined with the above
equation results in the total reaction rate for dissolved
CO2 (rsco2):
Equation 10.12
where:
(yb
r1)
= rate of generation of sco2t by the hydrolysis of vss
(yd
r2)
= rate of generation of sco2t by the growth of
methanogens
gvol
= gas constant for CO2 in L/mole
The mass transfer of CO2 between the liquid and gas
phases (r6) is negative when CO2 is transferred
from the liquid to the gas phase.
Ammonia
Ammonia (snh) is assumed to be produced only at the
hydrolysis/acidification stage. Ammonia (snh) is utilized by
the methanogenic organisms for growth. The rate of generation of
free ammonia (rsnh) is modelled as:
Equation 10.13
where:
(yf
r1)
= rate of generation of ammonia by hydrolysis/acidification
-r2/yg
= rate of consumption of ammonia due to growth of methanogenic
organisms.
The chemical equilibria and the pH calculation used for ammonia
(snh) and ammonium ion (snhi) are based on
Hydromantis' pH Model..
Model Parameters
This section of the chapter discusses the various model parameters
and inputs that the user would encounter when using this model. The
Parameters menu is shown in Figure
10‑3
Figure 10‑3
– Parameters Menu for the Anaerobic Digester
Physical and Operational
Parameters
The physical parameters for the Mantis2 digester model are the
volume of the liquid phase or maximum volume (vm),
the effective volume fraction (effvol), the headspace
volume (vgas), the total gas pressure (itpcon),
and the digester temperature (temp). The effective
volume fraction accounts for differences in the designed and actual
digester volume (due to mixing, inert solids, precipitation,
etc).
Figure 10‑4 - Physical Parameters
The Operational parameters shown in Figure
10‑5 are exclusively related to the control of the pumped
flow. These parameters are similar to the control parameters used
in other models (e.g., CSTR reactor).
Figure 10‑5
–Operational Parameters
Influent and Effluent
Parameters
These menu items
are found under Parameters sub-menu items Influent
and Effluent.
The Influent sub-menu,
shown in Figure 10‑6, allows the user to define
influent parameters that are exclusive to the basic digester
model: soluble total CO2 (sco2t), toxic
substance concentration (stox), methanogens
concentration (xmh) and net cations (strong
bases)(sz).
The Effluent sub-menu, shown in Figure 10‑7,
allows the user to define special components to the effluent of the
anaerobic digester: inert soluble COD (si) and the
fraction of the effluent VSS that is inert
(frinert).
Figure 10‑6
- Influent Parameters (Basic Digester Model)
Figure 10‑7
– Effluent Parameters (Basic Digester Model)
pH Solver Set up
Parameters
These parameters,
found under the Parameters sub-menu item pH Solver Set
up (Figure 10‑8) define the initial pH
value (ph), the pH boundaries: minimum (lowph), and
maximum (highph) values, and the required pH accuracy
(errorph). Parameters for the search routine of the pH are
presented in this sub-menu.
Figure 10‑8
- pH Solver Set up
Kinetic and Stoichiometric
Parameters
The kinetic
parameters (shown in Figure 10‑9) are found under
Parameters sub-menu item kinetic. The maximum
specific growth rate for methanogens (mumh) and the
rate constant for hydrolysis of vss (kco) are defined
for 35 degrees Celsius and corrected by the model using the
Arrhenius equation with the temperature coefficients indicated at
the bottom of this menu (tmumh & tkco). The rest
of the kinetic parameters in the basic model are not
temperature-dependent.
Figure 10‑9
- Kinetic Parameters
The stoichiometric parameters (Figure 10‑10) are
found under Parameters sub-menu item Stoichiometric.
The first set of stoichiometric parameters consists of conversion
factors: particulate COD (xcod) to vss ratio
(icvcon), BOD5 (bod) to BOD
ultimate (bodu) ratio
(fbodcon).
Conversion factors that
are unique to the digester basic model are the mass acetic acid
to COD factor(ac2cod), the molecular weight of
fatty acids (mwfat) and the gas constant
(gvol).
Using these factors
and the parameters defined in the Effluent sub-menu, the
basic digester model establishes the values for the composite
parameters. These parameters will modify the stoichiometry of
the effluent stream.
Figure 10‑10
- Stoichiometric Parameters
The second set of stoichiometric parameters consists of yields. The
relative amounts of chemical components produced by the biological
reactions in the anaerobic digester are specified by these
yields.
Other
stoichiometric parameters in this sub-menu are the dissociation
constants used to calculate the ionized components in the pH model
incorporated in the basic model. The dissociation constant for
ammonium (kncon) is defined for 20 degrees Celsius and
is temperature sensitive, i.e., the model corrects it for
temperature changes using the following equation:
Equation 10.14
The rest of the dissociation constants are not corrected for
temperature changes in the basic model.
Gas transfer
parameters are also defined on this form. The mass transfer
coefficient for carbon dioxide gas (sco2) between the liquid
and gas phases (KLaco2) is not corrected by temperature in
the basic model. Henry's law constant for carbon dioxide
(henryco2) is also included in this item and is not
corrected for temperature.
Anaerobic Digestion Model
#1 (ADM1)
Anaerobic Digestion
Model #1 (ADM1) (Batstone et al., 2002) is
implemented in GPS-X according to the ADM1 COST Benchmark (Rosen
and Jeppsson, 2002), with the following changes:
·
Several differential equations that describe the acid-base
equilibrium of the system (equations for Sva-,
Sbu-, Spro-, Sac-,
Shco3-, and Snh3) have been converted to
algebraic equations as described in Table B.3 of Batstone et
al. (2002). These processes are very fast and
contribute to the stiffness of the system of differential
equations.
·
The differential equation for Sh2 has been converted to
an algebraic equation to improve the simulation speed. This
process is very fast and contributes to the stiffness of the system
of differential equations. This approach is described by
Rosen et al. (2005).
These changes substantially increase the solution speed of ADM1 in
GPS-X and allow an integration algorithm other than Gear’s Stiff to
be used. Double Precision arithmetic should be used when
solving ADM1 in GPS-X
(accessed in
Options > Preferences > Build
tab)
The structured model includes five process steps including
disintegration, hydrolysis, acidogenesis, acetogenesis and
methanogenesis. The model uses 32 dynamic state variables, 6
acid-base kinetic processes, 19 biochemical processes, and 3
gas-liquid transfer processes.
A simplified ADM1 material flow diagram is shown in Figure
10‑11. For a full description of the model, the
reader is referred to Batstone et al,. (2002).
Implementation details are found in Rosen and Jeppsson
(2002).
Figure 10‑11
- Simplified ADM1 Material Flow Design
In GPS-X, the ADM1
model makes use of an ASM1 to ADM1 interface developed by Copp
et al. (2003). This interface allows ADM1 to be
used within a full-plant layout that uses ASM1 to model activated
sludge processes.
When using ADM1
there are two possible scenarios:
·
The influent stream is represented by an influent object.
This case is more difficult and requires that you specify the
influent in the influent object and in the Influent Form in the
digester object. See Figure 10‑12 for a
graphical description of this procedure. The section entitled
“ADM1 Model Set up Suggestions” gives suggestions on
how to characterize the influent.
·
The influent stream is an output stream from another object.
This case is simpler as the object that precedes the digester will
take care of the characterization of the stream itself. The
user still needs to specify the parameters found in the Influent
form of the digester object. See Figure 10‑13
for a graphical description of this procedure. The section
entitled “ADM1 Model Set up Suggestions” gives
suggestions on how to characterize the influent.
Figure 10‑12
- Method of Specifying the ADM1 Influent when the Influent Stream
is Represented by an Influent Object
Figure 10‑13
- Method of Specifying the ADM1 Influent when the Influent Stream
is an Output Stream from another Object
The following hints are useful for characterizing influent streams
to a digester object using the ADM1 model. Because the ADM1
model uses a different set of state variables than the ASM
activated sludge models, supplementary information has to be
supplied for the ASM1/ADM1 interface.
In typical simulation practice, you will not have all the
information you need, and will have to estimate the values for many
of the ADM1 influent parameters. These guidelines can be
useful for making sure that the estimates are used correctly.
The following suggestions are for the Carbon-Nitrogen Library
(CNLIB).
ADM1 State Variables Set on the ADM1 “Influent”
Parameter Menu
Table 10‑1 – ADM1 State Variables Set on
the ADM1 “Influent” Parameter Menu
Parameter
|
Suggestion
|
cations
|
Must be
measured or estimated or can be set to the value of the
Sic (inorganic carbon) state in ADM1.
|
anions
|
Must be
measured or estimated or can be set to the value of the
Sin (inorganic nitrogen) state in ADM1.
|
long chain
fatty acids
|
Must be
measured or estimated. It is part of Ss in the influent
object form.
|
total
valerate
|
Must be
measured or estimated. It is part of Ss in the influent
object form.
|
total
butyrate
|
Must be
measured or estimated. It is part of Ss in the influent
object form.
|
total
propionate
|
Must be
measured or estimated. It is part of Ss in the influent
object form.
|
total
acetate
|
Must be
measured or estimated. It is part of Ss in the influent object form.
|
hydrogen
gas
|
Set to
zero
|
methane
gas
|
Set to
zero
|
proteins
|
Must be
measured or estimated. It is part of Xs in the influent
object form.
|
sugar
degraders
|
Set to
zero
|
amino acid
degraders
|
Set to
zero
|
long chain
fatty acid degraders
|
Set to
zero
|
valerate
and butyrate degraders
|
Set to
zero
|
propionate
degraders
|
Set to
zero
|
acetate
degraders
|
Set to
zero
|
hydrogen
degraders
|
Set to
zero
|
ADM1 State Variables Set
in the Influent Object
When the influent
stream is represented by an influent object, the states influent
model should be used in the influent object. Some of the
states are ADM1 states and these need to be estimated. Some states
are not relevant for ADM1 and can be ignored. Some states form a
fraction of an ADM1 state and when added up need to correspond to
the values entered into the influent form of the ADM1 object.
More detailed descriptions are given below:
Table 10‑2 – ADM1 State Variables Set in
the Influent Object
Parameter
|
Suggestion
|
inert
inorganic suspended solids
|
Must be
measured or estimated. It is not an ADM1 state but will impact the
solids concentration.
|
soluble
inert organic material
|
Same as the
Si state in ADM1
|
readily
biodegradable substrate
|
Sum of
Monosaccharides, Amino acids, Long chain fatty acids, Total
valerate, Total butyrate, Total propionate, Total acetate, and
Carbohydrates states in ADM1. Make sure this value is the sum
of the values entered for these states in the ADM1 influent
stream.
|
particulate inert organic material
|
Same as
Xi state in ADM1.
|
slowly
biodegradable substrate
|
Sum of
composites, proteins, and lipids states in ADM1. Make sure this
value is the sum of the values entered for these states in the ADM1
influent stream.
|
active
heterotrophic biomass
|
Set to zero
unless you have an estimate of it.
|
active
autotrophic biomass
|
Set to zero
unless you have an estimate of it.
|
unbiodegradable particulates from cell decay
|
Becomes part
of Xi state in ADM1. Unless you know Xu, use
Xi above and set this to zero
|
internal
cell storage product
|
Not used.
|
dissolved
oxygen
|
Not in ADM1
but used by GPS-X to estimate the COD demand of the influent
stream. It is assumed that both oxygen and nitrate will be
reduced instantaneously upon entering the anaerobic environment, so
the total incoming COD is reduced by this incoming COD demand to
compensate for the reduction of the electron acceptors present in
the incoming stream.
|
free and
ionized ammonia
|
Sin state in ADM1 (inorganic nitrogen). Make sure
that correct units are used.
|
soluble
biodegradable organic nitrogen
|
(nitrogen
faction in amino acids = 0.098) amino acids concentration
|
particulate biodegradable organic nitrogen
|
Use the
following formulas:
Xnd = total TKN - Snd - Snh -
inxu*Xu -inxi*Xi
total TKN = Sin +
NiSi
+ Ni(Xi+Xu) +
NaaSaa
+ NxcXc
where:
Snd = conc. of soluble biodegradable organic
nitrogen
Snh = conc. of free and ionized ammonia
inxu = nitrogen fraction of unbiodegradable particulates
from cell decay (0.068 in Mantis, 0.06 in ASM1)
inxi = nitrogen fraction of particulate inerts (0.068 in
Mantis, 0.06 in ASM1)
Xu = conc. of unbiodegradable particulates from cell
decay
Ni = nitrogen fraction of Si and Xi
Si = conc. of soluble inerts
Xi = conc. of particulate inerts
Naa = nitrogen fraction of amino acids (0.098)
Saa = conc. of amino acids
Nxc = nitrogen fraction of composites (0.0376)
Xc = conc. of composites
|
nitrate and
nitrite
|
Not in ADM1 but
used by GPS-X to estimate the COD demand of the influent
stream. It is assumed that both oxygen and nitrate will be
reduced instantaneously upon entering the anaerobic environment, so
the total incoming COD is reduced by this incoming COD demand to
compensate for the reduction of the electron acceptors present in
the incoming stream.
|
dinitrogen
|
Set to zero.
|
alkalinity
|
Sic
state in ADM1 (inorganic carbon).
|
The objective of the aerobic digestion process is to stabilize and
reduce the mass of solids for disposal. In this process,
microorganisms consume their own protoplasm for energy; they are
assumed to be in the endogenous phase. This phase is accounted for
in the biological models associated with the CSTR object, but in
aerobic digestion, destruction of particulate inert organic
material also occurs.
The models associated with
the aerobic digester are given the name <model>dig,
where <model> is the associated activated sludge model (such
as asm1, etc.). The only difference between the activated sludge
model and its corresponding aerobic digester model is an added
first-order reaction. This additional reaction adjusts the
particulate inert organic concentration to account for the
destruction of the inert organics:
Equation 10.15
where:
rxi
= rate of reaction for particulate inert organics (xi)
(under Parameters – Kinetics menu of the anaerobic digester
object)
ki = inert
bioconversion rate
No loss of COD is
involved in this process, and no electron acceptor is utilized.
This destruction process converts particulate inert organics
(xi) to slowly biodegradable substrate (xs). The
slowly biodegradable substrate formed is then hydrolyzed, releasing
an equivalent amount of readily biodegradable COD.
The UASB/EGSB model
is available only in MANTIS2LIB. The Mantis2 model is used to model
the biological-chemical reactions in the reactor. Some of the
assumptions made in the development of model are listed below.
1.
The hydraulic regime in the UASB/EGSB is modeled as completely
mixed reactor.
2.
The substrate diffusion into the granule is assumed to be not
limiting and the reactions are modeled similar to the suspended
growth systems.
3.
The average granule properties are used to estimate the granule
settling velocity and bed expansion in the reactor.
4.
A semi-empirical model is used to estimate the solid concentration
distribution above the bed.
The simplified model structure is a first attempt to dynamically
model the UASB//EGSB reactors for practical engineering
problems.
Some of the important inputs and outputs of the model are described
in following sections.
Reactor Parameters
Additional sets of input parameters are required for the UASB/EGSB
model to estimate the granule settling velocity, bed fluidization
and solid distribution profile above the sludge bed.
The Reactor
Parameters can be accessed from the Input
Parameters menu item (Figure 10‑14). The
Reactor Parameters form is as shown in Figure
10‑15.
Figure 10‑14
- Reactor Parameters for UASB/EGSB Reactor
Figure 10‑15
- Reactor Parameters Input Form
Fraction of
un-reacted flow-through solids
It is well known
that UASB/EGSB reactors are not very effective in treating the
suspended solids in the wastewater. A fraction of the solids in the
influent stream may just flow through the UASB/EGSB reactor without
getting adsorbed or reacted in the reactor. This parameter allows
users to specify the fraction of solids which will pass through the
reactors un-reacted. Although, this phenomenon is normally
observed, it is not very well quantified. The default value of this
parameter is set at 0.5. In actual situation, depending on the bed
expansion, the value of this parameter may vary.
Average granule
size
This parameter is
used to calculate the settling velocity of the granules in the
granular bed. The estimated settling velocity is also used in
estimating the bed expansion in the reactor. A default value of 2mm
is used.
Terminal
velocity reduction factor
The observed
settling velocity of the biological granules is found to be less
than the settling velocity estimation procedures. Normally a
reduction factor of 0.7-0.8 is suggested in the literature.
Water content in
granule
This parameter
defines the water content in a granule and is used to estimate the
density of the granule required in the settling velocity estimation
procedure.
Void ratio of
stationary granular bed
This parameter
defines the void volume in the granular bed under no flow
conditions.
Depth of
transition zone
The UASB/EGSB model
uses a semi-empirical model to decide the solid profile above the
expanded granule bed. In the present model it is assumed that there
exists a transition zone above the sludge bed in which the solid
concentration changes from the concentration in the bed to a
fraction of concentration at the end of transition zone. The depth
of this transition zone will affect the solid profile and solid
capture efficiency in the reactor.
Ratio of SS in
transition zone to sludge bed
As described above,
this parameter represents the ratio of solid concentration at the
transition zone boundary to the solid concentration in the granule
bed.
Fraction of
non-settleable solids
This parameter
signifies the fraction of finer particles in the bed which are
non-settleable. This ratio is expressed with respect to the solid
concentration in the reactor.
Solid recovery
efficiency of gas solid separator
The solid
distribution curve above the granule bed is used to estimate the
solids concentration reaching the gas solid separator at the top
UASB/EGSB reactors. Depending on the efficiency of the gas solid
separator the concentration of solids escaping in effluent is
estimated.
CHAPTER
11
The proportional splitters in GPS-X can be set up for constant flow
split, variable split based on a timer, or variable split based on
flow. In the Parameters > Splitter Set up form you
can select the appropriate splitting mode. They are described
below:
Constant
When using this splitting mode, the split fractions do not change
(as entered under the Constant section), or are read from a
file, or are controlled by any other method (controller tool or
customization).
Timer based
When using this splitting mode, the split fractions, as entered
under the Constant section, will be rotated through the
connection points according to the specified time interval.
Flow based
When using this splitting mode, the split fractions, as entered
under the Constant section, will be rotated through the
connection points according to the specified volume. Once the
specified volume has passed through the splitter input, the
fractions will be moved to the next output connection.
A typical example of using this
functionality is a multiple train SBR plant, where the influent has
to be sent to a different SBR tank either by timer control or by
discharged volume.
A control splitter can be used to split flow into a set flow rate
and an overflow. This object is useful for setting flow bypass
controls.
The pumped flow
rate is set in the Parameters > Pumped Flow menu.
If the incoming flow is equal to or less than the pumped
flow rate, all of the flow will exit through the pump
connection point. If the incoming flow is greater than the
pumped flow rate, the excess flow will exit through the
overflow connection point.
An automatic PID
controller can be used in this object to control another variable
in GPS-X with the pumped flow rate.
The noreact model for Pumping Station allows user to use
continuous or intermittent pumping based on height or volume based
control.
Model Data Input
The input form is
as shown in Figure 11‑1:
Figure 11‑1 -
Pumping Station Menu
Continuous pumping - this variable controls whether
continuous pumping or intermittent pumping is used.
Average daily pump flow rate - this flow rate is used for
continuous pumping rate and also in the calculation of steady state
calculations when intermittent pumping is selected.
Control Type - the option gets activated when continuous
pumping is OFF. User may select from volume or height based control
options.
Low volume threshold for pump - this variable defines the
volume in the tank below which the pump operates at low threshold
pumping capacity. The option is activated when Volume Based control
is selected.
High volume threshold for pump - this variable defines the
volume in the tank above which the pump operates at maximum
threshold pumping capacity. The option is activated when Volume
Based control is selected.
Low height threshold for pump - this variable defines the
height in the tank below which the pump operates at low threshold
pumping capacity. The option is activated when Height Based control
is selected.
High height threshold for pump - this variable defines the
height in the tank above which the pump operates at maximum
threshold pumping capacity. The option is activated when Height
Based control is selected.
Minimum pump capacity - this is the capacity at which the
pump operates when the volume in the tank is below the low volume.
The option is activated when continuous pumping is OFF.
Maximum pump capacity - this is the capacity at which the
pump operates when the volume in the tank is above the high volume.
The option is activated when continuous pumping is OFF.
One model is
available for this object (empiric).
In this zero-volume model, the
soluble state variables are not affected, but solids are
partitioned into one of two streams based on a user-defined solids
separation factor. Flow can be pumped from either of the effluent
connection pipes and the difference between the pumped flow and
influent flow is diverted to the other effluent connection. Solids
concentrations in the two effluent streams are calculated through a
mass balance based on flows and the separation factor. An example
calculation is given below:
Influent
Flow
|
=
|
100
m3/d
|
Influent
xbh
|
=
|
1000
g/m3
|
|
|
|
Mass Influent
xbh
|
=
|
100 x 100 =
100,000 g/d
|
|
|
|
Pumped Flow
(user-defined)
|
=
|
10
m3/d – concentrate
|
Filter
Efficiency (user-defined)
|
=
|
0.90
|
|
|
|
Filtrate
Flow
|
=
|
90
m3/d
|
Filtrate xbh
Concentration
|
=
|
100
g/m3
|
Filtrate xbh
Mass
|
=
|
90 x 100 =
9,000 g/d
|
|
|
|
Concentrate
xbh Mass
|
=
|
100,000 –
9,000 = 91,000 g/d
|
Concentrate
xbh Concentration
|
=
|
91,000 / 10 =
9,100 g/m3
|
In the nonreactive noreact model, the effect of dilution on
wastewater components is described, but all reaction rates are set
to zero (no biological reactions are occurring). The tank has
aeration terms (for modelling of pre-aeration) and a built-in pump,
and both aeration and pumped flow can be controlled with feedback
controllers (P, PI, PID) analogous to the ones found in aeration
tanks and settler/clarifier objects.
This model can be used to
simulate an equalization tank either off-line or in-line, as the
volume is variable. In an off-line or variable volume mode (flow
equalization), the pump connection should be used with the proper
controller set up, while for concentration equalization in fixed
volume tanks the overflow connection is best used.
Reactions can be added by
double-selecting the model (noreact) and editing the rate
equations of the individual components.
In a sludge
pretreatment process, the particulate organic compounds are
converted into soluble organic compounds. The degree of
solubilization normally depends on the intensity of the treatment.
There are many different types of sludge pre-treatment processes.
These processes can mainly be divided into thermal, mechanical and
chemical treatment. Thermal, mechanical disintegration, ultrasound,
microwave, alkaline treatment ozonation, Fenton etc. are some of
the treatment which have been normally applied. For thermal and
mechanical treatments, the main operational parameter affecting the
degree of solubilization is the specific energy input, while for
chemical treatment it is the specific chemical dose. The specific
energy input and specific chemical dose are normally expressed in
terms of per unit of solids. As the mechanisms in sludge
pretreatment processes are normally too complex to describe through
a mechanistic model, empirical approach of modelling is normally
adopted. For general engineering purpose, this approach is
sufficient to evaluate different options and help in decision
making.
The general sludge
pre-treatment process model in GPS-X includes the following
transformations.
1.
The model describes the relationship between degree of
solubilization to the operational parameters like specific energy
input or specific chemical dose using a solubilization saturation
curve.
2.
The model allows conversion of the non-biodegradable organics into
biodegradable organics.
3.
The model allows for loss of during the treatment
The sludge
pretreatment model is available in MANTIS2LIB. Three main
transformations are described in the model.
1.
Inactivation of bacterial biomass
The sludge
pre-treatment normally results in bacterial inactivation. The
bacterial inactivation may happen due to disintegration of the cell
wall and release of soluble organic products. The biological
inactivation will results in production of soluble and particulate
organic products. Depending on the nature of treatment as
part of the COD may be lost due to oxidation.
2.
Conversion of Inert Organics (Xi and Xu)
The inert
particulate COD is reduced during the sludge pre-treatment. A part
of the particulate inert COD is assumed to convert to the slowly
biodegradable COD. The particulate inert COD also solubilize and
results in soluble biodegradable and soluble inert COD. Depending
on the treatment a part of the inert organic COD may be lost.
3.
Conversion of slowly biodegradable COD (Xs)
The particulate slowly and very slowly biodegradable COD is
solubilized and results in the soluble biodegradable substrate.
The implemented
model is an input-output model in which all the input states are
instantaneously converted to the output states. It is assumed that
the retention time in the chemical-mechanical disintegration
reactors are small.
The In-line
Chemical Dosage object (Figure 11‑2) can be used
to simulate chemical addition for soluble phosphorus removal and
coagulation/flocculation of soluble and colloidal components. Two
chemical precipitation models chemeq and metaladd are
available in GPS-X. The metaladd model is available in all
the libraries having soluble phosphorus as a state variable. The
chemeq model is not available in MANTIS2LIB. The
metaladd model considers the effect of pH and solubility
product in the metal precipitation reactions.
Figure 11‑2
- In-line Chemical Dosage Object
Chemeq Precipitation
Model
The P-removal model
is available only in the CNP library. The model is based on the
stoichiometry of the chemical precipitation of soluble
phosphorus. The model is set up to simulate the addition of
one of four possible chemicals:
1.
alum;
2.
ferric compounds;
3.
ferrous compounds; and
4.
other user defined metal compounds.
In each case, the dosage is expressed in terms of mass of metal
ion. For example, for dosing alum, the dosage is expressed in
terms of mass of aluminum ion.
As an example, the basic reaction of phosphorus
with aluminum is shown in Equation 11.1.
Equation 11.1
Therefore, one mole of aluminum ion reacts with one mole of
phosphate or, on a mass basis; one gram of aluminum ion
(Al3+) reacts with 1.148 grams of phosphorus
(P). The dosage that is required is the mass of aluminum ion
that is added rather than the mass of alum, which may be in
different forms (hydrated, etc.). This method of calculating the
stoichiometric requirements for alum and the other chemicals avoids
the issue of different compounds and their molecular weights. The
phosphorus stoichiometric coefficients for alum and the other
chemicals are already defined, but could be modified for modelling
another chemical that is not predefined.
Although the
stoichiometric ratios are used as the basis of the precipitation
model, the complexities of wastewater chemistry, including the
effects of pH, alkalinity and other elements, means that the
stoichiometric ratio is only the maximum achievable phosphorus
removal in presence of a large overdose of chemical. In practice, a
significantly lower removal is achieved because all of the
chemicals added are not available for phosphorus addition. However,
the available fraction, and therefore the required dosage, must be
estimated or calibrated from site to site and specified in the
model.
With a large
overdose of metal ion to wastewater, most of the phosphorus would
be removed; however, there would still be trace amounts of
phosphorus. To model this phenomenon, a saturation function was
incorporated in the model. The amount of phosphorus removed (or
soluble organic component, si and ss, that would
flocculate) becomes a function of the phosphorus concentration (or
soluble organic component concentration). The stoichiometric amount
of removal is only achieved when the phosphorus concentration is
large (with respect to the half-saturation coefficient).
Removal is less than the maximum stoichiometric amount when the
phosphorus concentration is small (with respect to the
half-saturation coefficient). This phenomenon is depicted in
Figure 11‑3.
Figure 11‑3 -
Removal as a Function of P Concentration
The mass of
chemical precipitant formed is also calculated and added to the
particulate inert inorganic component (xii). The
stoichiometric amount of chemical sludge produced is based on the
stoichiometry of the basic reaction. For example, in Equation
11.1, the mass of chemical precipitant (AlPO4)
is 4.52 g per 1g of aluminum ion reacted.
The model provides
the user with two dosage methods: mass flow based and
flow proportional. The former requires the user to specify a
continuous flow rate for the chemical, while the latter allows the
user to specify a flow rate of chemical per unit flow of liquid
entering the basin. The dosage rate can be set up as a manipulated
variable to maintain the phosphorus at a specified set point at any
location downstream of the chemical addition.
In addition to
precipitating the phosphorus, the chemicals flocculate some of the
colloidal organics into particulate matter. In the model this is
simulated as a conversion of soluble inert COD and soluble
substrate into particulate inerts and particulate substrate. The
extent of flocculation will depend on the half-saturation constants
for each species. These constants need calibration from plant to
plant as the flocculation model is empirical.
Here is a detailed
explanation of the constants of the Chemical Dosage menu
item:
Number of
chemicals: Currently four chemicals (alum,
ferrous and ferric, plus a user defined) are available.
Dosage mode . Mass
flow based: in mass/time units.
Flow
proportional: in mass/volume units.
The following
stoichiometry parameters (for all compounds) are found under the
More… button, and are described in Table
11‑1.
Table 11‑1 -
Chemeq Parameters
Parameter
|
Description
|
P
precipitation stoichiometry
|
gP/gMe precipitated stoichiometrically (if metal
is in excess)
|
soluble
inert COD flocculation stoichiometry
|
gCOD/gMe soluble inert COD converted to
particulate inert COD (si to xi)
|
soluble
substrate flocculation stoichiometry
|
gCOD/gMe soluble substrate converted to
particulate substrate (ss to xs)
|
metal
hydroxide stoichiometry
|
g of MeOH formed per g of metal added
|
P
precipitation half-saturation coefficient
|
gP/m3 concentration at which half the
Metal available for this reaction will be used to precipitate
phosphorus
|
soluble
inert flocculation half-saturation coefficient
|
gCOD/m3 concentration at which half
the metal available for this reaction will be used to flocculate
soluble inerts
|
soluble
substrate flocculation half-saturation
coefficient
|
gCOD/m3 concentration at which half
the metal available for this reaction will be used to flocculate
soluble substrate
|
The stoichiometry for inorganic inert solids production from P
precipitation that was available as a parameter in previous
versions of this model (under the name chemdos) is now
calculated from known stoichiometry.
In addition, the chemeq model is written as an
equilibrium model, meaning that all precipitation and
flocculation is assumed to happen instantaneously, and goes to
completion within the flow-through of the In-line Chemical Dosage
object.
All nitrogen and
phosphorus fractions of COD components are recalculated for mass
balance after the precipitation has taken place.
Dosage Controller
The chemeq
model contains a P-removal dosage controller, which is available in
the Input Parameters > Chemical Dosage menu (see
Figure 11‑4).
Figure 11‑4 – Dosage
Controller Parameters
The controller
allows users to specify a desired effluent P concentration, and
back-calculates the required metal ion dosage, based on the
stoichiometry settings.
Metaladd
Precipitation/Coagulation Model
The metaladd
model is a semi-mechanistic model to assess the removal of soluble
phosphorus due to metal addition. Following key assumptions are
made in the model development.
1.
The metal species exists in the form of Me(OH)3. All the
other soluble and insoluble species of metal are not modeled. Thus,
the model results outside the pH range of 6-8 may not be
applicable.
2.
Solubility product of the metal phosphates is used in estimating
the amount of expected precipitates.
3.
The solubility product is based on the PO43-
ion which is estimated based on the solution pH and dissociation
coefficient of H3PO4.
4.
The removal mechanism is assumed to be instantaneous, therefore no
volume of the reactor is considered.
Following mass balance equations may be written for different
phosphorus components across the In-line Chemical Dosing
object.
Mass Balance
Equation for Me(OH)3 Concentration
Equation 11.2
Mass Balance
Equation for Soluble Phosphorus Concentration
Equation 11.3
Mass Balance
Equation for MePO4 Concentration
Equation 11.4
The equation describing the relationship among the metal ion,
phosphate ion and the solubility product is as below:
Equation 11.5
where:
Xme(oh)3,o
= metal hydroxide concentration at outlet, gMe/m3
Xme(oh)3,in
= metal hydroxide concentration at inlet, gMe /m3
Pme
= metal dose, gMe /m3
a
= stoichiometry conversion factor, gMe(OH)3 /g Me
Xmepo4,formed
= concentration of formed precipitate, g
MePO4/m3
b
= stoichiometry conversion factor, gMe(OH)3/ g
MePO4
c
= stoichiometry conversion factor, g P/g MePO4
a
= dissociation factor for estimating PO43-
from total orthophosphate concentration
The set of above equations are solved simultaneously to calculate
the soluble phosphorus removal in the in-line chemical dosing
object. The main calibration parameter in the model is the
solubility product of the metal phosphates. The default values are
provided based on the evaluation of reported values in literature.
The values however may need to be adjusted to account for the
differences in wastewater characteristics, mixing regimes and other
parameters which affect the P-precipitation process.
In addition to P-precipitation, the metaladd model also
models the removal of colloidal COD (scol), soluble inert
COD (si) and soluble organic nitrogen (snd) through
flocculation. Following assumptions are used in the formulation of
this empirical flocculation model.
1.
The metal dose to flocculate unit soluble component is a function
of the soluble component concentration.
2.
The metal dose required per unit soluble component removed is
expressed as an exponential function with minimum bound.
The equation for metal dose per unit ss removed can be
expressed as below:
Equation 11.6
where:
F =
metal consumed per unit soluble component at soluble component
concentration C, g Me/g component
C =
Concentration of soluble component, g/m3
Fmax = maximum
metal consumption per unit soluble component, g Me/g component
Fmin = minimum
metal consumption per unit soluble component, g Me/g component
ka
= affinity factor, m3/g
The shape of the
curve represented by the above equation is shown in Figure
11‑5:
Figure 11‑5
– Curve Showing Relationship between the Soluble Components
Concentration, and Require Metal Dose for Unit Soluble Component
Removal
The integration of
curve in Figure 11‑5 between the limits of Sin
(concentration of soluble component in influent) and So
(concentration of soluble component in effluent) results in the
following expression. For a given metal dose and inlet
concentration, the equation is solved to estimate the So,
concentration of soluble component in effluent.
Equation 11.7
Equation 11.8
where:
Pme = metal dose,
gMe/m3
Sin =
concentration of soluble component in influent, g/m3
So
= concentration of soluble component in effluent,
g/m3
There are three parameters Fmax,
Fmin and ka used in the model.
The parameter Fmax may be interpreted as the
amount of metal required to remove the said component at very low
concentrations. It is general observation that the amount of metal
required at lower concentration of soluble component is much higher
than the metal amount required when the concentration of soluble
component is higher. The parameter of Fmin, can
be interpreted as the amount of metal required to remove the said
component at high concentrations. The parameter of
ka signifies the transition between the
Fmin and Fmax. Higher values of
ka will result in a very steep transition between
the Fmin and Fmax, essentially
improving the removal efficiencies. Conversely, lower values of
ka will result in a mild transition and thus
decreasing the removal efficiencies.
It is important to highlight here regarding the fate of the removed
soluble components. As, the flocculation process does not cause any
mass loss, the removed soluble compounds are just converted into
their particulate state. The si, scol and snd
are just transformed to xi, xs and xns states
respectively.
It should be noted
that the value of these parameters depends on the environmental
conditions temperature, pH, mixing, and wastewater characteristics.
Therefore, it is recommended that adequate calibration is performed
to find the appropriate values for specific cases.
The model
associated with this object is a simple empirical model where grit
production is directly proportional to the flow coming in the grit
chamber. The amount of grit produced (in grams per m3 of
influent) is user specified.
The struvite
precipitation reactor model represents an up-flow granular bed
reactor with recycle. The model uses a bed fluidization model
to estimate the granular bed height and the solid concentration in
the effluent stream. An internal recycle is provided to achieve bed
fluidization velocity required by the user. A typical layout
showing the application of struvite reactor is as shown in
Figure 11‑6. Some typical outputs from the struvite
precipitation model are as shown in Figure 11‑7and
Figure 11‑8.
Figure 11‑6 -
Typical Application of Struvite Recovery Reactor
Figure 11‑7 -
Typical Process Outputs for Struvite Reactor
Figure 11‑8 -
Typical Outputs for Solid Bed Volume and Expansion for Struvite
Recovery Reactor
Empiric Model
The empiric model associated with this object computes the
fraction of surviving E. coli based on the detention time in
the disinfection unit. The state variables are not affected by
this model. The following equation is used as the model for
computing the surviving fraction:
Equation 11.9
where:
survecoli
= fraction of surviving E.coli
kill
= rate of kill (m3/g/min)
ccl2
= chlorine dosage (g/m3)
tde
= detention time (computed from volume/flow rate)
The chlorine dosage, rate of kill, and
volumeof the unit are specified as operational
parameters.
Chlorination Model
Chlorine is widely
used for disinfection due to its cost effectiveness and high
efficiency. In addition to its disinfection ability, chlorine can
also reduce ammonia in the wastewater. However, a major concern of
using chlorine as a disinfectant is the formation of toxic
by-products such as Tri-Halo-Methane (THM) and Halo-Acetic Acid
(HAA). The by-products formation can be extremely difficult to
control and monitor due to hypochlorous acid’s high reactivity with
various compounds present in wastewater. The optimization of
chlorine dosages is often required to meet the disinfection
requirement; to minimize the by-product formation and to meet the
free chlorine discharge limits in the wastewater effluent to avoid
damage to aquatic environment.
An empirical
chlorination model is developed which can model the stoichiometric
reactions between chlorine and ammonia to estimated combined and
free chlorine residuals. The model also uses empirical
relationships to estimate production THM, and HAA formation.
Based on the available chlorine dosage, the disinfection
effectiveness is also estimated.
Instantaneous Chlorine
Demand
The following
expression is used to estimate instantaneous chlorine demand
required by the organic compounds in water (Dharmarajah et al.,
1991). The parameters of TOC and UV254 are used as representative
organic compounds leading the instantaneous chlorine demand.
Equation 11.10
Where:
Cinst - instantaneous chlorine
demand, mgCl2/L
- empirical constant 1, -
- empirical constant 2, -
- empirical constant 3,-
- empirical constant 4,-
sTOC - soluble total organic carbon, mg/L
-absorbance at 254nm, cm-1
The
UV254 is calculated by using a linear relationship
between UV254 and the the soluble inert organics
compounds (si) in wastewater.
Chlorine Decay
A first order rate
equation is used to model the decay of chlorine in wastewater.
Equation 11.11
Where:
Cavailable
- available chlorine, mgCl2/L
Cdose
- chlorine dose, mgCl2/L
kinact
- decay constant, 1/hr
t
- detention time (t10), hr
In the above
expression, the chlorine dose is adjusted for any chlorine lost due
to instantaneous demand. The net available chlorine is used in the
chlorine-ammonia reaction and in by-product formation.
Chlorine-Ammonia
Reactions
A typical
breakpoint chlorination curve is as shown in Figure
11‑9. When chlorine is added to water containing ammonia,
mono-chloramine is formed (zone 1). After all of the ammonia
has reacted, the free chlorine reacts with the mono-chloramines to
form di-chloramine and to nitrogen gas (zone II). The free
chlorine residual starts to appear after breakpoint chlorination
(zone 3).
Figure 11‑9 -
Breakpoint Chlorination [3]
Zone 1 occurs when
the Cl2/NH3-N ratio is below 5.07 and
involves the formation of chloroorganic and mono-chloramine
compounds. The formation of di-chloramine and nitrogen tri-chloride
is assumed to be negligible.
Zone 2 occurs when
the Cl2/NH3-N ratio is above 5.07 and below
7.58. Chloramines and chloroorganic compounds are destroyed in this
zone and the chloramines concentration becomes the lowest at
breakpoint. In theory, breakpoint occurs when the
Cl2/NH3-N ratio is 7.58, all the ammonia
nitrogen is consumed by chlorine and free chlorine becomes
available in zone 3. As the actual stoichiometric ratio may get
affected by the water composition, the default stoichiometric ratio
may be changed by the user in the model. The chlorination model was
developed to reproduce the break point curve shown in Figure
11‑9. Using the above reaction scheme, the model calculates
free chlorine (chlorine in HOCl and OCl- form),
combined chlorine (chlorine in combined form with ammonia) and
total chlorine (free + combined chlorine) in the system. Zone 3 is
where the by-products formation is assumed to dominate as free
chlorine concentration increases.
In zone 1, the
ammonia gets converted chloroamine, hence no loss of nitrogen is
considered. In zone 2 part of the chloroamine gets fully oxidized
leading the nitrogen loss as nitrogen gas.
By-product Formation-
Trihalomethane (TTHM)
The formation of
Trihalomethane (THM) occurs in the presence of hypochlorous acid
(free chlorine). The expression used to determine the
concentration of total trihalomethane (TTHM) formation is based on
batch experimental work conducted by Amy et al. (1987) and is
presented below.
Equation 11.12
Where:
CTTHM = total trihalomethane concentration
(mmole/L)
UV254 = water UV absorbance at 254 nm; 1 cm path length
(cm-1)
Cl2 = chlorine
residual (mgCl2/L)
t = reaction
time in the batch reactor (hr)
T = temperature
(oC)
pH = pH after chlorine reaction
(-)
Br = bromide concentration in
raw water (mg/L)
The reaction time,
t, is considered to be represented by the t50 value for
a unit process. If chloramines are present, the TTHM formation has
been observed to be reduced compared to when only chlorine is
present. As a result, a factor of 0.2 times the general TTHM
equation is applied when chloramines are present (USEPA, 1992).
By-product
Formation- Haloacetic Acid (HAA)
Haloacetic Acid
(HAA) formation is based on the AWWA Technical Advisory Workgroup’s
(TAW) work. The data were obtained from a series of batch
experiments, 96 hours in duration, using water from eight
facilities. The work is presented in USEPA (1992), and is
used as the default model. The expressions used to determine
HAA formation are presented below and were obtained from the USEPA
(1992).
Equation 11.13
Equation 11.14
Equation 11.15
Where:
CMCAA = monochloroacetic acid (ug/L)
CDCAA = dichloroacetic acid (ug/L)
CTCAA = trichloroacetic acid (ug/L)
sTOC = soluble TOC, mg/L
The Ai,j are
the empirical model constants. It is assumed that the
t50 value represents the reactor time t. If
chloramines are present, a factor of 0.2 is applied to each of the
individual HAA species concentrations, similar to the approach used
for TTHMS when chloramines are present.
Disinfection Model
The following
equation is used for estimating the log inactivation for
disinfection.
Equation 11.16
Equation 11.17
Equation 11.18
Where:
t
= time in batch reactor (min)
V
= volume of batch reactor ()
Q
= flow rate of wastewater ()
chlorine dose = chlorine present in the batch reactor
(g/m3)
ratio
=
baffling factor (short circuiting factor)
pH
= pH after chlorine reaction (-)
T
= temperature (oC)
WERF Model (werfuv) with
UV Disinfection
The werfuv
model is based upon an empirical model presented in a 1995 WERF
report comparing UV irradiation to chlorination (Water Environment
Research Foundation, 1995). The model estimates the effluent
coliform density:
Equation 11.19
where:
N =
effluent coliform density (MPN/100mL)
D
= UV dose (mW·s/cm3)
n
= empirical coefficient related to UV dose (unitless)
f
= empirical water quality factor (unitless)
The empirical water
quality factor, f, was determined from regression to be
described by:
Equation 11.20
where:
SS = suspended
solids concentration (mg/L)
UFT = unfiltered UV
transmittance at 253.7nm (fraction)
A,a,b = empirical
coefficients (unitless)
Four models are
associated with this object: asce, empiric, simple
and press. The models all simulate solids
separation, but do so via different models that have been
calibrated for different dewatering technologies.
Figure 11‑10 - Dewatering Object
Models
ASCE Model
The
asce model
is based on a paper by the Task Committee on Belt Filter Presses
for the American Society of Civil Engineers (ASCE, 1988). Based on
this work, which involved information collected as part of a survey
through the United States, a relationship was developed between the
feed solids and cake concentration. The primary sludge
fraction (i.e. percent of sludge for processing originating
from primary sludge) is the main parameter in determining the cake
solids from the unit. The solids capture is used to
determine the centrate solids concentration and the developed
relationship is used to determine the cake solids. A material
balance is used to determine the two output flows.
Empiric Model
The empiric
dewatering model is used when modellers wish to specify an
underflow solids concentration and flow rate, or specify a
concentration removal efficiency along with either the underflow
flow rate or solids concentration.
Figure 11‑11 - Operational Menu for
Empiric Model
In the
Operational Menu, users set the method for setting solids
removal by selecting from the drop down menu. The relevant
fields (pumped flow, underflow solids concentration, removal
efficiency) are highlighted.
Once two of the
three above options have been specified, the model performs a mass
and flow balance (similar to the simple model below) to
determine the flow and concentrations for the cake and filtrate
streams.
Simple Model
In the
simple model, the
user inputs two operational parameters: Solids capture (%),
and Polymer dosage (g/m3). From these two
parameters, the dewatering parameters (under Parameters >
Calibration) and the input solids concentration
(Xi), the cake (Xs) and
centrate (Xo) solids concentrations are
calculated:
Equation 11.21
The optimal polymer
dosage and sludge treatability are used to define the performance
of the unit. The minimal sludge concentration for processing is a
limit concentration under which the unit will not operate. The cake
dry material content is a function of the treatability and the
expected content from easy to treat sludge at optimal polymer
dosage, hard to treat sludge at optimal, and no polymer
dosage.
Using the following mass balance,
the output flow rates (Qo and
Qs) can be computed:
Equation 11.22
Press Model
The press
model is based on the
simple model but has, in addition, a parameter for the
filter press surface area and a maximum load on the
press. These two parameters are used to model the reduced
efficiency of dewatering as the loading on the press increases.
The dewatering object should not be confused with
the thickener object. The thickener is configured like a
nonreactive settling tank. The Sedimentation Models
are described in CHAPTER 8.
Differential Model
In certain solid-liquid separation devices (centrifuges, vortex
separator etc.), the inorganic particulate matter may be removed at
higher efficiency than the organic particulate matter due to
difference in the density. Further, depending on the nature of
aggregation/particle size of particulate matter, there may be
differences in the relative removal efficiencies of different
organic particulate matters.
In some of the
sludge reduction systems (i.e. Cannibal Process), it is suggested
that the inert organic material is preferentially removed by the
solid-liquid separation device in the process. To model such
processes, an empirical solid-liquid separation model
(differential) is implemented in dewatering unit process
considering differential removal of particulates depending on the
characteristics of each particulate state variable.
In the differential
settling model, the particulate matter concentration in the
effluent is calculated based on the following equation:
Equation 11.23
where:
Xj,o
= concentration of particulate state j in effluent, mg/L
Xj,
in
= concentration of particulate state j in influent, mg/L
aeff
= removal efficiency
The concentration in the underflow (concentrated) stream is
estimated by making a mass balance around the object.
Equation 11.24
where:
Xj,
un = concentration of
particulate state j in underflow, mg/L
Qin
= influent flow rate, m3/d
Qo
= effluent flow rate, m3/d
Qun
= underflow rate, m3/d
The underflow rate, Qun, is
specified by the user as a constant value or as a fraction of the
influent flow rate. The overflow rate is estimated by using
the flow balance equation as below.
Equation 11.25
The soluble compounds in the influent are mapped to the effluent
and underflow streams without any transformation.
The disc and belt
microscreen objects also contain the empiric solids separation
model, as described above, but with different default removal
rates. The disc microscreen object is found in the “Tertiary
Treatment” object group, and the belt microscreen object is found
in the “Preliminary Treatment” object group. Both objects
have only the empiric solids separation model.
Figure 11‑12 – Belt and Disc Microscreen
Objects
The hydrocyclone
model found in the Biosolids Treatment section of the unit process
table is a simple solids separation unit. It uses the same
differential model as discussed above in the dewatering
object. The solids capture rates for the various particulate
components have been calibrated for typical performance. In
particular, the solids capture rate for Anammox biomass is higher
than that of the other biomass types, reflecting the heavier nature
of the biomass which allows it to settle out preferentially.
The High-Rate unit
process object is found in the “Biosolids Treatment” section of the
unit process table. It contains the highrate solids
separation model.
Highrate Model
Background
The highrate model is based on a vortex separator and retention
treatment basin (RTB) model developed by Gall et al.
(1997) and Schraa et al. (2004) for the
high-rate treatment of wet weather flows such as combined sewer
overflows (CSOs). The model has been applied to primary clarifiers,
chemically enhanced primary clarifiers, RTBs, vortex separators,
continuous backwash filters, and ballasted flocculation (e.g.
Actiflo) and sludge recycle (e.g. Densadeg) systems.
The highrate model
contains a hydraulic routing component and a settleability
component. The hydraulic component of the model consists of
algorithms to achieve dynamic volumetric and mass balances.
The settleability component of the model includes a relationship
between settleability and influent TSS concentration, and a
relationship between settleability and the surface overflow rate
(SOR). The settleability model for high-rate retention
treatment basins is given below:
Equation 11.26
where:
E
=concentration based removal efficiency (%)
Eu
= ultimate settleable solids fraction (at high influent TSS)
CTSS = parameter
that controls the exponential shape of the removal efficiency vs.
Influent TSS relationship
TSSinf = influent total suspended
solids concentration (mg/L)
SOR = surface overflow rate
(m/h)
K =
surface overflow rate a 0.5 Eu
In this model Eu, CTSS, and
K are parameters that must be calibrated using measured
data. Once calibrated, the model can be used to predict the
removal efficiency achieved in a high-rate solids separation
process, in terms of total suspended solids (TSS), given the
influent TSS and the surface overflow rate (SOR).
The default settleability model parameters provided in GPS-X are
based on data taken from studies conducted by the University of
Windsor on a pilot-scale RTB operated at the Lou Romano Water
Reclamation Plant (LRWRP) in Windsor, Ontario, Canada
(Li et al., 2004) that used high doses of
cationic polymer as the sole coagulant. A 3‑dimensional
scatter plot of the pilot plant settleability data is given in
Figure 11‑13. Tests that did not use polymer
were removed from the data set. The circles represent data
points and the lines or stems show where the points lie in the
plane.
At a constant influent TSS, the removal efficiency was observed to
decrease with increasing surface overflow rate. At a constant
surface overflow rate, the removal efficiency increased
exponentially with increasing influent TSS and leveled off at TSS
values above approximately 500 mg/L.
Figure 11‑13
- Plot of TSS Removal Effciency vs. Influent TSS and SOR (Pilot
Data from U. of Windsor RTB)
The settleability
model was calibrated to the entire set of pilot data shown in
Figure 11‑13, using the optimizer tool
in GPS-X and the sum of squares objective function (i.e. least
squares parameter estimation). The calibrated parameter
values are given below:
Eu
= 0.86 (ultimate settleable solids fraction at high influent
TSS)
CTSS = 82
mg/L (parameter that controls the exponential shape of the removal
efficiency vs. influent TSS relationship)
K = 93
m/h (surface overflow rate at 0.5Eu)
A 3-dimensional
plot of the calibrated settleability model is shown in Figure
11‑14. The model shows the same trends as the
measured pilot data shown in Figure 11‑13. To
check the performance of the calibrated settleability model,
dynamic simulations were performed using the full RTB model and the
measured influent TSS concentrations from the pilot tests.
The simulation results showed reasonable agreement with the
measured pilot study results given the estimated measurement
error.
Figure 11‑14
- Plot of Calibrated Settleability Model
Physical Setup Form
The Physical
Setup form is accessed by right-clicking on the dewatering
object and selecting Input Parameters >
Physical. Figure 11‑15 shows the GPS-X
form. The parameters are discussed below:
·
Number of trains in operation: Specifies the number of
high-rate treatment process trains.
·
Surface area per train: Specifies the surface area of each
train (assuming all trains have the same surface area). The
default SI units are m2.
Figure 11‑15
- Operational Menu for High-Rate Treatment Model
Operational Setup
Form
The Operational Setup form is accessed by right-clicking on
the dewatering object and selecting Input
Parameters > Operational. Figure
11‑16 shows the GPS-X form. The parameters are
discussed in the next section.
Figure 11‑16
– Operational Menu for High-Rate Treatment Model
High-Rate Treatment Model
Sub-Section
Equation Shape
– specifies the settleability model to be used. The user
can select the following options: Constant TSS Removal
Efficiency, Exponential Function of Influent TSS, Switching
Function-Based SOR Relationship, Combined Equation
·
Constant TSS Removal Efficiency: Uses the maximum
settleable solids fraction to calculate the effluent TSS based on
the influent TSS. For example, using E =
100Eu as the settleability model.
·
Exponential Function of Influent TSS: Uses an
exponential-based expression to calculate the effluent TSS based on
the influent TSS. For example, using the following equation as the
settleability model. This model increases TSS removal
efficiency as the influent TSS increases
Equation 11.27
·
Switching Function-Based SOR Relationship: Uses a
switching function, based on the surface overflow rate (SOR) to
calculate effluent TSS. For example, using the following equation
as the settleability model. This model reduces the removal
efficiency at higher SORs.
Equation 11.28
·
Combined Equation: Uses the entire settleability
equation to calculate the effluent TSS, based on the influent TSS
and the SOR. For example, using Equation 11.26 (shown
below) as the settleability model.
Equation 11.29
Maximum
Settleable Solids Fraction: Specifies a maximum
concentration-based TSS removal efficiency fraction. This is
the highest TSS removal efficiency that can be achieved. The
quantity is dimensionless (-) but can also be entered as a
percentage
(%).Exponential
constant for TSS dependence: This parameter controls the
exponential shape of the removal efficiency vs. influent TSS
relationship. Increasing this parameter will decrease the TSS-based
removal efficiency. The default SI units are mg/L.
Surface overflow rate at half the maximum TSS removal
efficiency: This parameter is the half-saturation coefficient
in the removal efficiency vs. SOR switching function-based
relationship. Increasing this parameter will increase the TSS
removal efficiency. The default SI units are in m/d.
Operational Parameters
Sub-Section
Underflow
Setup: This drop-down menu allows the user to select how
the underflow rate and concentration are calculated. (Underflow
Rate, Underflow Concentration)
·
Underflow Rate: The underflow rate is specified and
GPS-X calculates the underflow concentration using a mass
balance.
·
Underflow Concentration: The underflow TSS
concentration is specified and GPS-X calculates the underflow using
a mass balance.
Underflow
Rate: This is the underflow rate used when the Underflow
Rate option is selected. The default SI units are
m3/d.
Underflow Concentration: This is the underflow TSS
concentration used when the Underflow Concentration option is
selected. The default SI units are mg/L.
Summary of the High-Rate
Treatment Model Output Forms
The Outputs
menu is accessed by right-clicking on the dewatering object and
selecting Output Variables. Figure 11‑17 shows
the Output Variables menu options.
Figure 11‑17 -
Output Variables Menu in High-Rate Treatment Model
The Output
Variables menu options that are unique to the high-rate
treatment model are discussed below:
Solids Removal
Efficiency: The solids removal efficiency form is shown
in Figure 11‑18. The user can plot or track the solids
removal efficiency.
High-Rate
Treatment Performance Parameters: The high-rate treatment
process loading rates form is shown in Figure 11‑19.
The user can display the surface overflow rate and/or the solids
loading rate.
Figure 11‑18
- Solids Removal Efficiency Output Variable Form
Figure 11‑19
- High-Rate Treatment Performance Parameters Output Variables
Form
The advanced
oxidation unit process is found in the Tertiary Treatment panel in
the unit process table. The model is an empirical model for
modeling soluble COD removal from wastewater using oxidising agents
like ozone, hydrogen peroxide etc.. The user specified oxidant dose
and oxidant effectiveness factor are used to oxidize soluble
organic compounds like soluble inert COD, colloidal COD, readily
biodegradable COD, acetic acid COD, propionic acid COD and methanol
COD. The nitrogen and phosphorus fractions associated with the
soluble inert COD are hydrolysed, thus increasing the concentration
of ortho-phosphorus and ammonia-N concentration.
The input for
maximum COD removal efficiency is used to control the residual COD
in effluent at chemical overdose. The influent flow and chemical
dose are used to estimate the mass of chemical required for
treatment.
The Black
Box object contains three models: Empiric, Pipe,
and Interchange
Empiric Model
The empiric
model contains a number of empirical equations that can be set up
to change the value of the BOD, TKN and the suspended solids
between the input connection point and the output connection
point.
The available regression equations
are:
1)
constant
2)
proportional
3)
linear
4)
quadratic
5)
exponential
6)
power
7)
user-defined
After choosing the
correlation type, the user specifies what variable is the
independent variable (either influent flow, influent concentration
or both). Finally, the user must supply the parameters in the
regression equation. Equation 11.30 and
Equation 11.31 are typical regression equations that
can be set up using the Black Box object.
One independent
variables (solids), with proportional regression
Equation 11.30
Two independent
variables (flow and concentration), with linear regression
Equation 11.31
Interchange Model
There are two options available to specify the state variable
mapping: From menu, and From custom macro:
From custom macro: State variable mapping is specified
in the custom macro file in the interchange model file (accessible
by clicking on the interchange option from the model
menu). More advanced mapping calculations can be performed in
the custom macro (see Figure 11‑20). For details on how to
write custom calculations, see the Customizing GPS-X
chapter in the User’s Guide.
Figure 11‑20
– Custom Interchange Macro
The pipe model can be used to simulate the lag effect caused by
significant residence times in pipes or channels between unit
process objects. By default, GPS-X does not simulate any
travel time between objects.
The pipe model is a
plug-flow tank model without any biological reactions.
Advanced Pump Energy Model
The energy consumed in a pumping system depends on the pumped flow
rate, pump head and pump efficiency. For a give pumped flow rate,
the energy consumption will depend on pump head and pump
efficiency. The required pump head for a pumping system has
comprises of static head and dynamic head loss in the pipes, valves
and fittings. While static head is independent of the pumped flow
rate, the dynamic pump head is a function of pumped flow rate and
piping system. The pump efficiency is also a function of the pump
flow rate and normally available from the pump characteristic
curve. The simple pump energy models applied to variable flow
conditions typically use a constant head and constant efficiency,
leading to inaccurate estimation of pumping requirement. These
models are also not suitable to compare energy performance of pumps
having different pump characteristics curves. Therefore, in
situations where large variation in the pumped flow rates are
expected, better energy consumption estimates can be made giving
adequate attention to pump characteristics curve.
The advanced pump energy model in GPS-X allows user to dynamically
estimate the pump head and pump efficiency under variable pump flow
conditions using the pump characteristics curves. Two pump models
are implemented 1) Fixed speed pump model and 2) Variable speed
pump model.
Fixed Speed Pump Model
System Curve
Characteristics
The pump model requires that a system curve is defined for the
pumping system. In the GPS-X model following equation is used to
define the system curve.
Equation 11.32
Equation 11.33
Where:
= static system head, m
=
static dynamic head, m
=
dynamic head-loss coefficient, -
= flow rate in the system, -
The model requires
an input value for hstatic, the static head of the system. In
addition to this, a point (Q, hdynamic) on the dynamic head curve
needs to be specified by the user. This point on the curve is
estimate the dynamic head-loss coefficient (K). The GPS-X input
screen for system curve characteristics is as shown below in
Figure 11‑21 and Error!
Reference source not found.:
Figure 11‑21
- Inputs for System Curve Definition – Static Head
Figure 11‑22 - Inputs for System Curve
Definition - Dynamic Head
Pump Curve Characteristics
The pump model requires an input of pump characteristics curve. The
user can provide the pump characteristics curve for a specific
speed of the pump by setting pump flow rate, pump head and pump
efficiency for a selected number of points on the pump curve. An
example is shown in the table below:
Table 11‑2 - Sample Pump
Characteristics
Points on Pump
Curve
|
Pump
Flowrate
|
Pump
Head
|
Pump
Efficiency
|
( - )
|
(m3/hr)
|
(m)
|
(%)
|
Point
1
|
0
|
22.7
|
0.0
|
Point
2
|
250
|
21.8
|
19.1
|
Point
3
|
500
|
20.7
|
20.0
|
Point
4
|
750
|
19.8
|
37.3
|
Point
5
|
1000
|
18.7
|
63.6
|
Point
6
|
1250
|
17.6
|
72.7
|
Point
7
|
1500
|
16.5
|
78.2
|
Point
8
|
1750
|
15.6
|
82.7
|
Point
9
|
2000
|
14.5
|
84.5
|
Point
10
|
2250
|
13.6
|
86.4
|
Point
11
|
2500
|
12.4
|
86.4
|
Point
12
|
2750
|
11.3
|
85.5
|
Point
13
|
3000
|
10.0
|
82.7
|
Point
14
|
3250
|
8.5
|
77.3
|
Point
15
|
3500
|
6.9
|
69.1
|
For any given flow rate, the model uses pump characteristics curve
and system head loss properties to calculate the system operating
point (Figure 11‑23). Based on the pump operating point, the
model calculate pump efficiency, number of pump and energy
consumption in pumping.
Figure 11‑23
- Pump operating point based on pump and system curves
Variable Speed Pump
In this model, user can run the pump in a variable speed mode and
estimate the energy saving that may be achieved by converting a
fixed speed pump to a variable speed pump for a given system
conditions. An ON and OFF switch is provided in model for the
variable speed pump. In addition to above user inputs required for
the fixed speed pump, additional inputs for the minimum and maximum
speed of the pump are required. The variable speed model is built
with an energy optimization algorithm in which the model finds a
pump speed at which consumption of pumping energy is minimized. The
typical outputs for the variable speed pumps are:
1)
Number of pumps required to be operated
2)
speed of the pumps
3)
efficiency of the pump
4)
system head, pump head
The following affinity equations are used to calculate the flow
rate, pump head and power requirement at different speed of the
pump.
Affinity Law #1:
Affinity Law #2:
Affinity Law #3:
where,
Q1 = flow rate at N1 speed,
m3/d
Q2 = flow rate at N2 speed,
m3/d
H1 = pump head at N1 speed, m
H2 = pump head at N2 speed, m
P1 = power absorbed at N1 speed, kW
P2 = power absorbed at N1 speed, kW
The change in the pump curve at different pump speeds are as shown
in Figure 11‑24.
.
Figure 11‑24 - Pump curves at different pump
speeds
In the variable speed pumping, a VFD power efficiency factor is
introduced to take into account the power. The equation for
calculating different power consumptions are as shown below.
Phydraulic = ρ·g·Q·H
Pshaft = Phydraulic /
ηpump
Pmotor = Pshaft /
ηmotor
Pinput = Pmotor /
ηVFD
where,
P = power, kW
ρ = water density, kg/m3
g = gravity acceleration, m/s2
Q = flow rate, m3/d
H = total head, m
ηpump = pump hydraulic efficiency, -
ηmotor = motor efficiency, -
ηVFD = VFD (variable frequency drive) efficiency, -
Figure 11‑25
- Pump Characteristics Curve - Pump Speed for Pump Curve
Figure 11‑26
- Pump Characteristic Curve Inputs
Typical Model Outputs
Based on the input data of system head loss, pump characteristics
and a given flow rate, the model find the operating point for the
pump. It estimates the number of pumps required to pump the flow
and efficiency of the pump at the operating point.
A typical output for a pump system pumping a diurnal flow is shown
in Figure 11‑27.
Figure 11‑27
- Typical Output from a Fixed Pump Speed
It can be seen that when the pump flow rate increases to more than
the pumping capacity of a single pump, two pumps become
operational, each pumping half of the total flow rate [Figure
11‑27 A]. At this point since the flow rate are decreased
for each pump, the change in the pump operating point leads to
reduced pump efficiency [Figure 11‑27 C]. This
reduced pump efficiency increases the energy consumption
significantly for the pumping system. During the model calculation,
head used in the energy calculation is the greater of the system
head and pump head Figure 11‑27 C]. In general, the pump
head should always be higher than the system head for proper
pumping. This condition should always be checked by the user. If
the pump head is lower than the system head, an alternative pump
having more appropriate pump characteristics curve should be used.
In this particular example, constant speed is used throughout the
simulation.
Typical Model Outputs
The typical outputs from the variable speed pump model are as shown
in Figure 11‑28. Following observations may be made
by comparing the outputs of variable speed pump and fixed speed
pump.
1)
The pump head follows the
required system head more closely [Figure
11‑28 B] than that in
the fixed speed pump system. Synchronization of the pump head with
the system head results in reducing energy consumption.
2)
The increase in the speed of
the pump follows the increase in the flow rate. The optimum speed
of the pump is estimated to achieve the lowest power
consumption.
3)
The pump efficiency for the
variable speed system 0.84 – 0.88 as against the efficiency
variation in the range of 0.77 – 0.88 for the fixed speed
system.
4)
The power consumption for the
variable speed pump varies from 62-132 kW while for fixed speed
system the observed range was 100-180 kW.
It is clear from the above analysis that the variable speed pumps
can offer significant energy saving in a pumping system where flow
variations are large.
Figure 11‑28
- Typical Outputs for Variable Speed Pump
The building object
does not have any effect on the treatment of liquid or solids
streams. It is used for the estimation of energy requirements
of buildings on the treatment plant site. Users can place as
many building objects as needed on the GPS-X layout. The building
object calculates the energy usage for HVAC, lighting,
refrigeration, and miscellaneous in the building. Based on the
energy usage, the operational cost for the building is
calculated.
HAVAC
Heating, cooling,
and HVAC energy requirement for the building is estimated using
conductive heat transfer equation as shown below:
where:
= heat
transferred per second ()
= area ()
=
temperature difference)
= thermal resistance parameter ()
= performance factor (-)
HAVAC energy can be
estimated in GPS-X by specifying temperatures, physical and heat
resistance parameters of the wall, HAVAC efficiency as stated by
the manufacturer,
Lighting
Light energy usage
for the building is calculated by specifying the quantity and
wattage of the lighting device.
Refrigeration
Refrigeration
energy usage is estimated based on conductive heat transfer through
the walls, frequency of usage (air changed in the cooler), and
refrigerated material load. In GPS-X, the user specifies the
temperature, physical and heat resistance parameters of the cooler,
cooler performance as stated by the manufacturer, refrigerated load
characteristics, and usage frequency.
Other
Miscellaneous
Other miscellaneous
energy required for the building can be specified under this
section. The input can be specified by entering a power value or
entering a fraction of the total building power usage.
Energy requirements
and cost estimations for the building are shown in the Operating
Cost output menu, as shown in Figure
11‑29 below. Note that the energy requirement
for the building is included in the miscellaneous category for the
plant-wide energy demand and operating cost calculations.
Figure 11‑29
– Operating Cost Output Menu
Water Quality Index (WQI)
model
Effluent Quality
Index model in effluent unit process uses the user specified
concentration limits and compound specific weight factors to
estimate 1) time duration for which a plant is under violation with
respect to specified concentration limits 2) estimates the
instantaneous and moving average Effluent Quality Index and 3)
estimate the net and moving average Effluent Quality index.
The WQI is
representative of the effluent pollution load to a receiving water
body and is estimated using the expression provided in COST
simulation benchmark (EC, 2002).
Where:
– Instantaneous effluent flow rate, m3/d
n – number of compounds in EQI
estimation (-)
– weight factor of
compound in EQI (-)
–instantaneous concentration of compound in
EQI estimation
T – time period for
moving average calculation
The net WQI, which
is defined as the weighted pollution load over and above the
violation concentrations, is also calculated in the model as
below.
Where,
– violation concentration of compound in EQI
estimation
By default, a total
of 9 compounds including TSS, COD, BOD, NH3-N, TKN, TN, NOx-N,
Ortho-Phosphorus, and total phosphorus are used in the EQI
calculation. User can exclude contribution from any of these
compounds by setting the weight factor of that compound to
zero.
The cumulative time
of violation for a specific parameter is estimated by the
integrating the time during which the plant is out of compliance
with respect to that parameter.
The cumulative time
of violation is divided by the total time to estimate the time
fraction in violation.
CHAPTER 12
This chapter
discusses the models found in the Tools and Process Control
objects. The toolbox object is found in the Tools
section of the unit process table. The Process Control
section of the unit process table contains several objects.
Each model is described in separate sections below.
The sampler model is found in the toolbox object, and will
sample any stream either at regular time intervals (i.e., take a
sample every `x' hours) or in proportion to the flow rate of that
stream (take a sample every `x' m3 of flow). It reports
either daily or total simulation run average values.
The sampling tool
parameters are accessed by selecting Parameters > Set
up in the Process Data menu. These parameters are
detailed below:
Label of sample
stream: This is the label of the stream being
sampled. All of the state and composite variables in this stream
are sampled.
Sampling mode ( Off /
Flow Proportional / Regular Time Intervals ): This
switch controls how the stream is sampled. The stream can be
sampled in proportion to the flow or at regular time intervals.
Sampling period
( Daily / Total Run ): This is the time period used when
calculating the averages of the sampled values. The averages can be
calculated on a daily basis or for the entire run.
Take a sample
every (m3): This parameter controls how often the stream is
sampled when the sampling mode is set to Flow
Proportional.
Take a sample every (hour):
This parameter controls how often the stream is sampled when the
sampling mode is set to Regular Time Intervals.
The ph tool
is found in the Toolbox object in some libraries, and allows the
user to estimate the pH of a wastewater stream in the layout, or do
pH estimation directly from parameters entered into the menu.
These two options are available in the pH Model Set up menu
(Figure 12‑1).
Figure 12‑1
- pH Model Set up Menu
The pH model operation menu item specifies whether the
calculation is a standalone pH estimation using the components
under the Wastewater Components (stand-alone model) header,
or uses the components from a specific flow point in the plant
layout.
If you select link pH estimation to label, the calculation
will use the concentrations of ammonia, nitrate, etc., from the
stream label specified in label of point where pH is to be
estimated. If you would like to estimate the pH at
several different points throughout the plant, place one Toolbox
object per pH calculation into the layout, and link each box to a
different point in the layout where pH is to be estimated.
The calculation is based on a model developed by Vavilin et
al. (1995). The model is simplified to include only
strong bases, strong acids, total ammonia, total acetate, and total
carbon dioxide. The model uses an iterative process, in combination
with the concentration of each component, to determine the pH
value.
The
relationship between the concentration of a component and the
concentration of its ions is related through a dissociation
constant (K). The concentration of hydrogen ion is used to
determine pH:
Equation 12.1
The initial pH identified on the pH solver set up control form is
used to calculate the initial concentration of hydrogen ion. An
example of this control form is shown in Figure 12‑1.
The default initial guess for pH is 7.
A second control form, shown in Figure
12‑2, is used to input various equilibrium and temperature
constants.
Figure 12‑2
– Component Control Form
These
concentrations are used in a series of five equilibrium equations
that form the basis for the pH model:
Equation 12.2
Equation 12.3
Equation 12.4
Equation 12.5
Equation 12.6
where:
Kn =
ammonium dissociation constant (mol/L)
Ka =
acetate dissociation constant (mol/L)
Kw = water
dissociation constant (mol/L)
Kc1 =
CO2 dissociation constant – step 1 (mol/L)
Kc2
CO2 dissociation constant – step 2 (mol/L)
The dissociation constants used in these equations can also be
input on the control form shown in Figure 12‑2.
The default dissociation constants are taken from the literature.
The ammonium dissociation constant is corrected for
temperature.
To properly reflect the presence/absence of added alkalinity (lime
addition, etc.) the alkalinity state variable salk is
assumed to be all carbonate alkalinity that is available (and
completely dissociated) for acid neutralization.
In the above equations, totalCO2 is the total
CO2 available (converted to equivalent charge).
Because the alkalinity is assumed to be in the form of completely
dissociated CaCO3, a 2+ charge cation is also assumed
present in the equilibrium system.
The above equations are combined
with the charge balance shown below:
Equation 12.7
An implicit
nonlinear equation solver is used to solve this system of equations
for [H+]. The calculated pH values are bounded by the low
bound and high bound parameters shown in Figure
12‑1.
Once the solution is complete and
the concentrations of the ionized components have been calculated,
the concentrations of the non-ionized components are calculated
from the speciation equation shown above.
The calculated pH and the
concentrations of the individual components can be displayed on an
output form similar to that shown in Figure
12‑3. This pH model is also used in the HPO (High
Purity Oxygen) object for pH estimation.
Figure 12‑3
– Sample pH Outputs
The lowpass model is found in the toolbox object, and can be
used for analog or digital filtering, or a combination of both. The
input signal is fed to a first-order analog filter, a first-order
digital filter, and a zero-order sample-and-hold. Any of these
components may be used or bypassed. This is done by setting the
filter type as Analog, Digital, or
Analog+Digital, and by setting sample and hold to
either ON or OFF. When the filter type is set
to OFF, both filters are bypassed and the final output is
either the input signal or the sampled input signal, depending on
the state of the sample and hold setting. A signal flow
diagram is shown in Figure 12‑4. In this figure, the
switches are set for a digital filter with sampling.
First-order linear filters
have one tuning parameter, the cutoff frequency ωc, which
corresponds to the corner frequency, i.e. the reciprocal of the
filter time constant. The amplitude ratio of a first-order system
at the corner frequency is 1/√2, with corresponding phase
lag φ=45 degrees’
With u(t) and
y(t) representing the input and output of a first-order
filter respectively, the filter models are:
Analog:
Equation 12.8
Digital:
Equation 12.9
Figure 12‑4
– Signal Flow Diagram for the lowpass Model
One menu of
parameters is defined for the lowpass model. It is accessed
from the Tools Process Data menu by selecting
Parameters > Filter Settings. The available
parameters are detailed below:
Input
signal: The cryptic name of the variable to be
filtered.
Filter type (Analog / Digital /
Analog+Digital / Off): The value of this parameter
determines whether the signal flows through the analog or the
digital filter, or both.
Analog filter
cutoff frequency: The corner frequency of the first-order
analog filter, or the reciprocal of its time constant.
Digital filter time
step: The value of the time interval used in the digital
filter model. It is also used as the sampling time when Sample
and hold is set to ON.
Digital filter cutoff
frequency: The corner frequency of the first-order
digital filter, or the reciprocal of its time constant.
Sample and hold (ON -
OFF): The value of this parameter determines if the
digital filter output is sampled and held for the sampling period
specified above.
Digital filtering without sampling
is useful if the filtered signal is to be used in a unit that
already implements a sample-and-hold on its input.
The ON/OFF
controller object is found in the Process Control panel.
On/off controllers are simple forms of feedback control. This type
of controller is used in a wide variety of common applications. For
example, household heating and cooling systems operate using on/off
control.
Using ON/OFF
Controllers
The ON/OFF
controller parameters are divided into two groups: control variable
parameters, and manipulated variable parameters. The control
variable parameters are accessed by selecting Parameters >
Control Variable from the Tools Process Data menu.
The manipulated variable parameters are accessed by selecting
Parameters > Manipulated Variable from the
Tools Process Data menu.
Control Variable
Parameters
Controller (ON – OFF): A value of ON indicates
that the controller is active; a value of OFFindicates that
the controller is not active (i.e., not adjusting the manipulated
variable). This should not be confused with the `on' and `off'
states of the on/off controller.
Controller sampling
time: This parameter determines the length of time
between controller executions. A high value (corresponding to a
long interval between executions) may be appropriate for slower
processes (e.g., MLSS control using the sludge wastage rate), while
other processes may require faster sampling frequencies.
Control variable (CV)
with label: The cryptic name of the variable to be
controlled.
Upper bound: The
upper bound on the controlled variable. If the value of the
controlled variable exceeds this value, then the value of the
manipulated variable will be set to its low setting.
Lower bound: The lower
bound on the controlled variable. If the value of the controlled
variable is less than this value, then the value of the manipulated
variable will be set to its high setting.
Manipulated Variable
Parameters
Manipulated
variable (MV) with label: The cryptic name of the
manipulated variable.
MV setting when CV is at low
limit: The value of the manipulated variable is set to
this value when the value of the controlled variable is less than
its lower bound.
MV setting when CV is at high
limit: The value of the manipulated variable is set to
this value when the value of the controlled variable exceeds its
upper bound.
MV initial value: The
value to assign to the manipulated variable until the controlled
variable crosses either its upper or lower bound.
The PID object
contains two models: PID, and PIDforward.
Regular PID controllers are widely used in industrial applications
because of their simplicity and good performance. PID is an acronym
for Proportional, Integral, Derivative, which
are the three modes of the PID controller. In continuous form, the
PID controller algorithm is given by:
With derivative kick protection OFF:
Equation 12.10
With derivative
kick protection ON:
Equation 12.11
where:
MV = manipulated
variable
CV = controlled
variable
E =
error (setpoint-controlled variable)
Kc
= proportional gain, a controller tuning constant
TI =
integral time, a controller tuning constant
TD =
derivative time, a controller tuning constant
I
= initialization constant
The individual
modes of the PID controller can be used individually or in pairs;
common combinations include P-only and PI control, in addition to
full PID control. The output of the PI controller, for example,
includes the first two terms in the PID controller expression
above; the output of any combination can be calculated similarly.
The proportional mode is given by:
Equation 12.12
The integral mode
is given by:
Equation 12.13
The derivative mode
is given by:
Derivative kick
protection OFF
Equation 12.14
Derivative kick
protection ON
Equation 12.15
Implementation in
GPS-X
Since the vast
majority of PID controllers are now implemented using digital
computers, the discrete form of the PID algorithm is used more
frequently than its continuous form. Accordingly, all controllers
in GPS-X are modelled in discrete form. Bounds on the
manipulated variable are frequently used to reflect limitations due
to physical equipment or safety. Therefore, the PID algorithms
implemented in GPS-X have the following forms, where
Dt represents the
controller execution interval.
PID in Velocity Form
With derivative
kick protection OFF:
Equation 12.16
With derivative
kick protection ON:
Equation 12.17
Equation 12.18
PID in Position Form
With derivative
kick protection OFF:
Equation 12.19
With derivative
kick protection ON
Equation 12.20
Equation 12.21
Position and Velocity
Forms
The velocity form
is obtained by differencing the position form of the PID
controller. When using the velocity form, one calculates the
change in the value of the manipulated variable that is to
be implemented, as opposed to the absolute value calculated
by the position form of the algorithm. The velocity form has
certain advantages over the position form; in particular, it is
naturally protected against reset or integral windup.
This condition occurs when a persistent, nonzero setpoint error
results in a large value of the integral mode of the PID, forcing
the control action to `saturate' (i.e. to maintain the manipulated
variable at its minimum or maximum value). Even when the setpoint
error returns to zero, the control action remains saturated when
the position form of the PID algorithm is used. When using the
velocity form, the control action can return within the control
range after one sampling period.
Derivative Kick
Protection
Derivative kick
protection is a commonly used modification to the standard PID
algorithm. The modification, which affects the derivative mode
only, prevents large `jumps' from occurring in the manipulated
variable as the result of setpoint changes. These jumps occur
because a discrete setpoint change introduces a discontinuity into
the error signal, resulting in an "infinite" derivative (in the
continuous case), or a large value in the difference error signal
(in the discrete case). Therefore, it is recommended that
derivative kick protection be used for all PID controllers, unless
the behaviour described above is specifically desired.
Derivative Filtering
Derivative
filtering is another strategy by which large jumps in the
manipulated variable can be prevented. Filtering reduces the effect
of sudden setpoint changes and high frequency noise present in the
error signal, which may produce large values of the error
derivative. When derivative filtering is selected, the signal fed
to the derivative mode of the PID (error or controlled variable) is
passed through a discrete first-order filter before being used in
the calculation of the derivative mode. The filter is tuned by a
single parameter, the cutoff frequency, which corresponds to the
filter's `corner frequency'. (i.e.‑the reciprocal of its time
constant)
PID Controller
Parameters
The parameters available for PID controllers are listed below.
These parameters are divided into two groups: control variable
parameters, and manipulated variable parameters. The control
variable parameters are accessed by selecting
Parameters > Control Variable from the
Tools Process Data menu. The manipulated variable parameters
are accessed by selecting Parameters > Manipulated
Variable from the Tools Process Data menu.
In addition to the PID controller
in the tools object, PID controllers for common variable pairings
are built into many of GPS-X's process objects. For example, a PID
controller for controlling the dissolved oxygen concentration is
available in the plug flow tank object. These controllers have
essentially the same parameters as the PID controller in the tools
object except that the manipulated variable and possibly the
controlled variable have already been defined.
Control Variable
Parameters
Controller (ON - OFF): This parameter determines
whether or not the PID controller is active.
Setpoint for control
variable: This is the value of the setpoint for the controlled
variable. The controller acts to keep the controlled variable at
the setpoint.
The following parameters are
accessed by clicking on the More... button in the Control
Variable form. Only the parameters in the Controller Set
up sub-section are discussed here. For a discussion on the
Controller Tuning sub-section see Tuning PID
Controllers.
Controller form (Full Position / Velocity): This parameter
is used to specify which form of the PID controller equation is
used (position or velocity). The default velocity form is
recommended because it has built-in protection against integral
windup.
Controller type (P / PI
/ PID): This parameter specifies the modes to be used in the
PID controller. Using the PI option is recommended in most
cases. When using the P option, the offset between the
controlled variable and the setpoint cannot be completely
eliminated.
Control variable with
label: This is the cryptic variable name for the
controlled variable including the appended stream label.
Controller sampling time: This is the length of time
between controller executions. If this parameter is too large the
control performance will be poor and possibly unstable.
Proportional
gain: This is the controller tuning constant,
Kc.
Integral time: This is the controller tuning constant,
TI.
Derivative
time: This is the controller tuning constant,
TD.
Controller
effect on controlled var - direct (ON - OFF): Since the value
of the proportional gain in GPS-X must be positive, this parameter
controls the `direction' of the controller. If a positive step
in the manipulated variable results in a positive change in the
controlled variable (at steady-state), i.e. the process gain is
positive, then this parameter should be ON(e.g. an increase
in air flow causes an increase in DO). Otherwise, it should
be OFF(e.g. an increase in wastage rate causes a
decrease in MLSS concentration).
Derivative kick
protection (ON - OFF): When this setting is ON the
manipulated variable is prevented from making large jumps in
response to setpoint changes. This parameter is only applicable
when the controller type is set to PID.
Derivative filtering (ON - OFF): The derivative filter is
used to filter the signal fed to the derivative mode of the PID
controller.
Cutoff frequency: This is
the derivative filter's corner frequency (i.e. reciprocal of its
time constant).
Manipulated Variable
Parameters
Manipulated variable w/o label: This is the
manipulated variable name without the stream label.
Label of manipulated
variable: This is the stream label of the manipulated
variable.
Value of manipulated
variable: This is the initial value of the manipulated
variable.
Minimum
value: This is the lower bound on the manipulated
variable.
Maximum value: This
is the upper bound on the manipulated variable.
Many different techniques exist for tuning PID controllers. While
an in-depth discussion of these techniques is beyond the scope of
this manual, some general guidelines are provided below.
A simple and conservative
approach to controller tuning is as follows:
·
Start with a small proportional gain and a large integral time; set
the derivative time to zero (0). As a rough rule of thumb:
o Use a
value of 0.5/Kp as a starting value for the
proportional gain (Kp is the steady-state process
gain, defined as the change in the controlled variable at
steady-state for a one-unit change in the manipulated
variable).
o Use a
value of 1.5t as a starting point for the integral time
(t is the time constant of the first-order response that
most closely approximates the controlled variable response to a
step in the manipulated variable).
·
Gradually increase the proportional gain until further increases
result in deterioration in control performance.
·
Gradually decrease the integral time until further decreases result
in deterioration in control performance.
·
Gradually increase the derivative time.
·
Fine-tune the parameters as needed.
A facility for tuning PID
controllers based on the Ciancone correlations for setpoint changes
is available for all PID controllers in GPS-X. The PID tuning
parameters calculated using the Ciancone correlations are based on
estimates of the process gain, dead time, and time constant,
obtained by a least-squares fit of the time series data collected
under tuning mode.
The controller
tuning parameters for the PID controller are accessed by selecting
Parameters > Control Variable > More... in the
Process Data menu of the tools object. Setting tuning
to ON starts the tuning mode. The controller sampling
time must be set to an appropriate value. The process should
have been brought to steady-state, or reasonably close to
steady-state, prior to the activation of the tuning mode. This may
be done with or without a PID controller currently ON. While
this mode is active, a step input is applied to the manipulated
variable, forcing the controlled variable to be
perturbed.
This mode may be
deactivated by manually setting Tuning to OFF, at
which point suggested PID tuning parameters will be displayed in
the simulation Log window, based on an analysis of the time
series data collected during the tuning operation. It is not
necessary to wait for a new steady-state before turning off the
tuning operation, but it may be desirable to do so. Tuning
parameters will also be suggested if the simulation terminates
while the tuning mode is ON.
The estimates of
the process parameters (i.e. process gain, dead time, and time
constant) may be tagged for display in the
Output Variables menu of the object or stream
associated with the controller.
The other settings which affect
the operation of the tuning mode are:
Fractional
step size: The size of the step to
be applied in the manipulated variable, specified as a fraction of
its initial value.
Time
of step: The interval of time
that will elapse between the activation of the tuning mode and the
application of a step in the manipulated variable.
Maximum
possible dead time: Upper bound on admissible
values of dead time in the first-order plus dead time model. This
may be used to reduce the time required to compute tuning
constants, or may be used to force a dead time of zero in the model
used to derive the tuning constants.
While operating in tuning
mode, time series data of the manipulated and controlled variables
are sampled (at each controller sampling time) and stored in arrays
of fixed maximum size. Data for 3000 sampling times can be stored.
If that capacity is to be exceeded, a warning will be displayed,
and a scrolling time window that will retain data for the last 3000
sampling times will be initialized. This should not be a limitation
in most cases, as this storage capacity is enough for data sampled
every 10 minutes over 20 days. The storage capacity is a global
setting labeled
controller tuning array size, found in General Data
> System > Parameters > miscellaneous. If
this value is increased, the layout must be recompiled for the
change to take effect, i.e., this value may not be changed inside a
scenario. For more information on tuning PID controllers,
consult Marlin (1995) or Perry and Chilton (1973).
The Ciancone correlations are presented and discussed in Marlin
(1995).
Users can select the PIDforward model in this PID controller
object, by right-clicking on the object and selecting Models
> pidforward. With a feedforward controller, an input
disturbance is measured and the controller adjusts the manipulated
variable to compensate for the disturbance before the controlled
variable deviates from its setpoint. Feedforward control is usually
combined with feedback control to retain the beneficial properties
of feedback.
The pidforward model of the Tools object combines a PID controller
for feedback and a lead/lag algorithm for feedforward control.
Lead/lag feedforward control is a simple feedforward strategy that
provides enough flexibility for most applications. The PID
controller algorithm is presented in the section entitled PID
Controllers above. In this section we present the lead/lag
algorithm.
A feedforward term may be
added to the feedback control action of the PID controller when
measurement of a process disturbance is available. The lead/lag
feedforward controller is represented in transfer function form
as:
Equation 12.22
Its time-domain
representation is given by:
Equation 12.23
where:
MVFF = additional control
action provided by feedforward control
D
= the measured disturbance
KFF
= the feedforward gain, a controller tuning parameter
q
= the feedforward
dead time, a controller tuning parameter
tld
= the feedforward lead time, a controller tuning parameter
tlg
= the feedforward lag time, a
controller tuning parameter
IFF
= the initialization constant
of the feedforward controller
When both the
process and disturbance dynamics are represented by first-order
plus dead time models, perfect feedforward control can in principle
be achieved if one:
·
Sets the feedforward gain KFF to the
disturbance-to-process gain ratio;
·
The dead time q to the
difference between disturbance and process dead times (possible
only when the process dead time is less than the disturbance dead
time);
·
tld
the lead time to the process time constant; and,
·
The lag time tlg
to the disturbance time constant.
Implementation in
GPS-X
The feedback
control action is calculated from a PID control algorithm as
described above, and the feedforward control action is added to it
as follows:
Feedforward/Feedback (PID in velocity form):
Equation 12.24
Feedforward/Feedback (PID in position form):
Equation 12.25
where:
ΔMVNFF
= the change in feedforward control action obtained from
discretizing the lead/lag algorithm and the superscript FF refers
to the feedback controller.
Feedforward Controller
Parameters
The parameters of the pidforward model are presented in
three menus: Manipulated Variable, Controlled
Variable, and Disturbance Variable. Feedback control
parameters are located in the Controlled Variable menu and
feedforward control parameters in the Disturbance Variable
menu.
The feedforward controller
parameters found in the Disturbance Variable form are
detailed as:
Feedforward
controller (ON- OFF): The feedforward
component of the controller is activated by this switch.
Disturbance variable
with label: The cryptic name of the disturbance variable
including the stream label.
Controller
sampling time: The length of time between controller
executions.
Feedforward gain: The
feedforward controller tuning parameter, KFF.
Leadtime: The
feedforward controller tuning parameter, tld.
Lag time: The
feedforward controller tuning parameter, tlg. The ratio of
lead to lag times should usually be kept below 2:1 to minimize the
effects of noise.
Dead time: The
feedforward controller tuning parameter, q.
Negative feedforward gain (ON
- OFF): This parameter controls the `direction' of
the controller. The `normal' sign of the feedforward gain is
negative. This occurs when both the process and disturbance gains
have the same sign, in which case this setting should be ON
(the default value). Otherwise, it should be OFF.
For information on
tuning Feedforward/Feedback controllers, consult Marlin
(1995).
The timer is used to schedule and control intermittent plant
operations. For example the timer control can be used to implement
intermittent aeration in an activated sludge tank. Another
application is the desludging of a primary settler.
The timer works in
cycles. At the start of each cycle the timer sets its manipulated
variable to a pre-defined ‘on’ value. The manipulated variable is
kept at this value for a user-defined fraction of the cycle. For
the remainder of the cycle, the manipulated variable is set to a
pre-defined ‘off’ value. This is ideally suited for intermittent
aeration because the controller can turn on the air flow at regular
intervals for a user defined length of time.
The timer control
parameters are accessed by selecting Parameters > Set
up in the Process Data menu. These parameters are detailed
below:
Controller (ON -
OFF): This switch activates the timer.
Manipulated variable with
label: The cryptic name of the manipulated variable
including the appended stream label.
Cycletime: The length
of one cycle.
Ontime in one
cycle: The length of time (starting at the beginning of
the cycle) that manipulated variable value is set to the `on' value
during a cycle.
Value of manipulated variable
when on: The `on' value of manipulated variable.
Value of manipulated variable
when off: The `off' value of manipulated variable.
The multivar
model is used in conjunction with the Advanced Control
Module. See the Advanced Control Module Manual for details on
using multivariable controllers in GPS‑X.
The Flow Timer
Model is a modification of the regular Timer model GPS-X. The Flow
Timer Model is developed to address the specific needs to define
daily intermittent flows using average daily flow rates. The main
difference between the timer and flow timer model is in the choice
of input variable. In the timer model, the input values of the
variables are instantaneous values while in the Flow Timer Model,
the flow variables are the average value in the cycle.
Model Data Input
The input form for
the Flow Timer Model is as shown in Figure 12‑5.
Figure 12‑5
- Flow Timer Menu
Controller – The controller switch is used to turn the flow
timer model ON/OFF. Flow will be controlled, only when the
controller is switched ON.
Manipulated variable with label – This input variable is
used to specify the cryptic variable name of the flow variable that
needs to be controlled.
Cycletime – This input variable is used to specify the total
duration of the cycle including the ON and OFF time.
Start time for ON cycle – This input variable controls the
start time of the cycle.
Ontime in one cycle – This input variable defines the
duration of ON cycle.
Average value of manipulated variable during cycle – This
input variable defines the average flow rate during the whole
cycle.
The default setup of the flow controller defines a cycle of total
length of 0.05d; the cycle remains OFF for 0.01d, gets ON at 0.01d
and remains ON until 0.04d. The flow is OFF from 0.4d-0.5d. This
cycle then repeats itself.
Some of the
operations at wastewater treatment plant follow a repeating
schedule. For example, the sludge wasting schedule may require
wasting on the weekdays and no wastage on weekends. Until now,
these repeating schedules were modeled by inputting the scheduled
inputs through a file input. This approach requires preparation of
file inputs for the duration of dynamic simulation. To simplify
modelling a repeating schedule, a new scheduler model is
implemented in the toolbox object of the GPS-X. The structure of
the scheduler model is generic and can be used to model cyclic
schedules of operation in a treatment plant. The scheduler model
can control flow rates, split ration and any other model input
variable.
The scheduler model
can be selected by right-clicking on the tool box object and
selecting scheduler. The scheduler model is useful when the
treatment plant is operated in a well-defined cycle containing many
different phases and in each phase a set of model variables is
manipulated in a well-defined manner. In the current scheduler
model, a set of 15 model variables can be manipulated to define
complex operational schedules in treatment plants.
Model Data Input
The Scheduler model
requires the following information:
1.
Number of phases in a cycle
2.
Time duration of each phase
3.
Cryptic names of model variables which are manipulated in a
cycle
4.
Value of each model variable in each phase of the cycle
The input data to
the model is set by using the input form of the scheduler model.
The input data form is as shown in Figure
12‑6. A brief explanation of each
input variable is given below.
Number of phase
in sequence: This variable defines
the number of distinctive phases in a cyclic-sequence. The default
value is set to 8.
Duration of each
phase: This variable defines the time
duration of each phase. The sum of the time duration of each phase
is equal to the total sequence (cycle) time.
Figure
12‑7 shows the input form for
setting the duration of each phase.
Cycle start phase: This defines the phase at which the
cycle starts. The value of the variable can be between 1 to number
of phases defined above.
Cryptic name of
model variables– This defines the cryptic names of the model
variables which are manipulated in each phase. By default, a
maximum of 10 model variables can be used to set the sequence.
All the variables are set to blank (nothing is manipulated in the
sequence). To include a model variable in the scheduler set-up, the
blank should be replaced by the full cryptic name of the variable
available in the layout. For example, to set the influent flow rate
for an influent object in the sequence, the blank variable can be
replaced by qconinf (qcon is the cryptic for influent
flow rate and inf is the stream id). If there are more than
one variable, they can be similarly defined.
TIP: To see the
cryptic name of a variable, you may point the cursor over the
variable description name in the data input window. The cryptic
variable name is displayed for few seconds in a small window.
Value of model
variable in each phase: The value of a model variable
during each phase can be set here. For example, to set the
values of control variable in each phase, click on the (…)
button to the right of the control variable to open the input value
form for model variables.
Figure 12‑6
- Input Form for the Scheduler Model
Figure 12‑7
- Input Form for Setting Duration of Each Phase
Figure 12‑8 - Input Form for Setting the
Value of Control Variable in Each Phase of a Sequence
CHAPTER
13
GPS-X contains operating cost models for most of the objects in the
process table. Each object can be set up to calculate costs for
energy, chemical dosage and sludge handling (as appropriate to that
object).
This chapter discusses the
structure of the operating cost models, and their use and
calibration.
For each of the wastewater unit process objects in GPS-X, a set of
operating costs has been assigned. These costs reflect the typical
operating costs associated with that particular unit process.
GPS-X can be used to
dynamically simulate these operating costs in the same way that it
dynamically simulates the wastewater unit process itself.
The different types of operating
costs modelled are:
·
Aeration energy cost
·
Pumping energy cost
·
Mixing energy cost
·
Heating energy cost
·
Other miscellaneous energy cost
·
Chemical dosage cost
·
Sludge handling cost
The following
sections described each cost in detail.
Aeration Energy Cost
An aeration energy
cost can be calculated for unit processes that have aeration, such
as the CSTR, plug flow tanks, SBR, etc. The amount of energy
required to supply the calculated level of aeration depends on
several factors, including blower/compressor efficiency, headloss,
etc., and the factors included in the oxygen transfer model (See
Modelling of Oxygen Transfer in Chapter 6).
GPS-X can simulate two different aeration methods - mechanical and
diffused air. If mechanical aeration is chosen, an aeration power
(kW) is entered. If diffused aeration is chosen, the power
requirement is calculated as shown in Equation 6.27
and Equation 6.28
in Chapter 6.
The wire power is multiplied by
the energy price ($/kWh), and then integrated over time to
determine the total cost for aeration energy during the
simulation
Blower Energy Cost
Energy required for
the pumping of water and wastewater is modelled in GPS-X using the
following equation:
Equation 13.1
Where:
= Blower power, kW
= Air temperature, ºC
= Air Flow rate, m3/d
= Air pressure at discharge, kPa
= Air pressure at the inlet, kPa
= Blower efficiency, -
= Atmospheric pressure, kPa
=
inlet pressure loss, kPa
= diffuser submergence, m
= head loss in piping and diffuser, kPa
The energy value is multiplied by the energy price ($/kWh), and
integrated over time to determine the total cost for aeration
energy during the simulation.
For blower energy costs,
the head value represents the sum of the actual head and the piping
headloss. There is no additional equipment headloss as with
aeration energy costs.
Mixing Energy Cost
The mixing energy cost model describes the use of energy for
mechanical mixing operations.
Ssers are required
to input the required mixing power usage per unit volume
(kW/m3). This value is multiplied by the volume of the
tank, and then multiplied by the energy price ($/kWh) and
integrated over time to determine the total cost for the
simulation.
Heating Energy Cost
The heating energy cost model describes the use of energy for
heating individual unit process operations such as digesters, or
activated sludge units in cold environments.
Users are required
to input a heating power usage per unit volume (kW/m3).
This value is multiplied by the energy price ($/kWh) and integrated
over time to determine the total cost for the simulation.
Other Miscellaneous Energy
Cost
The miscellaneous energy cost model describes the use of energy for
various mechanical operations such as gates, arms, rakes, moving
bridges, etc.
Due to the complex nature of
estimating these types of energy requirements, users are required
to input a flat rate of energy usage (kW). This value is multiplied
by the energy price ($/kWh) and integrated over time to determine
the total cost for the simulation.
Chemical Dosage Cost
Some unit process
objects in GPS-X allow for chemical addition as part of treatment
(some influents, equalization tank, DAF, dewatering units,
etc.).
In each case, the total chemical
dosage cost is:
Equation 13.2
where:
chemicalcostperday = chemical
dosage cost ($/d)
chemprice
= chemical price ($/kg)
chemdosagerate
= hydraulic head (kg/d)
This daily cost is integrated over time to determine a total cost
for the simulation.
Sludge Disposal
Several objects in GPS-X have connection points which represent
thickened or dewatered sludge (e.g. dewatering unit). A sludge
disposal cost can be associated with these flows to determine a
sludge handling cost.
The cost is
determined by multiplying the per-unit disposal cost by the rate of
sludge disposal:
Equation 13.3
where:
disposalcostperday =
sludge disposal cost ($/d)
disposalcost
= disposal price ($/m3)
sludgerate
= sludge disposal rate (m3/d)
This daily cost is
integrated over time to determine a total sludge disposal cost for
the simulation.
Operating cost
parameters are found in two places:
General Data
Forms: Right-click on an open spot in the layout.
These forms contain general operating cost parameters that apply to
the entire layout (i.e., energy prices and schedules).
Operating Cost
Parameter Forms (in each object): These forms contain
operating cost parameters that are specific to each object (i.e.,
headloss, pump efficiency, chemical cost).
General Operating Cost
Paramters
Operating cost
variables controlling the price of energy can be found in the
Layouts > General Data > System > Input Parameters
> Operating Cost Settings menu (See Figure
13‑1).
Figure 13‑1
– General Operating Cost Parameters Form
The price of energy
can be set in two different ways, by selecting from the Energy
Pricing menu:
·
Constant Price
·
Time-based Pricing (energy price varies throughout the day)
·
Seasonal Pricing
When Constant Price is selected, the energy price is set to
the value entered on the form. This value can be changed in a
scenario or by placing the price on an interactive controller, or
set constant for the entire simulation.
When Time-based Pricing is
selected, the energy price will cycle through a user-defined set of
prices on a user-defined schedule. The number of different energy
prices is unlimited. This mode can be used to simulate lower energy
costs during the night.
The Seasonal
Energy Pricing model allows the user to evaluate the cost of
power consumption under a dynamic price structure based on season,
weekday, weekend and hour of the day. The model is an extension of
the Time-Based Pricing model which only allows specification of
daily variation in energy price. The start day and month for each
season is required to be set by the user.
In seasonal pricing
model, two seasons of summer and winter are set each of which can
have different energy pricing structure. Three tiers of off-peak,
mid-peak and on-peak prices can be specified by the user.
In each season,
weekdays and weekends are allowed to have any hourly price
structure using the three tiers of pricing level.
The input form for
the seasonal price model is as shown below.
Figure 13‑2 – Input menu for seasonal
price model
In seasonal price
model, date and time set in the Simulation Setup menu is used to
determine the season and day name. This information is used to
determine the energy price based on the user inputs.
Object-Specific Operating
Cost Parameters
For all objects,
the operating cost model parameters can be found in the
Parameters > Operating Cost menu, and
its associated More... button. The values entered on these
forms are specific to each object. These objects can be tagged and
placed on controllers, etc., as with any other parameter.
Figure 13‑3 - Operating Cost Menu
The operating cost
model parameters shown in the Operating Cost parameters form
depend on the object and choice of model.
Similar to the
operating cost parameters, the associated output variables are
divided into general and object-specific groups.
General Operating Cost
Output Variables
Operating cost
output variables can be selected from the Layout >
General Data > System > Output Variables > Operating
Cost menu. These output variables include the energy price;
total energy, chemical and sludge disposal costs for the entire
layout; and the total operational cost for the layout (sum of the
total energy, chemical and sludge disposal costs).
Figure 13‑4 - Layout Operating Cost
Display Form (total for all objects)
Object-Specific Operating
Cost Output Variables
For all objects, operating cost model output variables can be
selected from the Output Variables > Operating
Cost menu by right clicking on the effluent connection
point.
The output variables available
include energy usage and sludge disposal rates as well as total
energy, chemical and disposal costs for that particular object. The
specific costs calculated (and available in this menu) depend on
the operating costs associated with the particular object and
model.
Figure 13‑5 - Object-Specific Operating
Cost Display Form
The operating cost models in GPS-X are designed to simulate typical
operating costs found in typical wastewater treatment facilities.
The models are populated with parameters that give costs for
typical plants.
To calibrate the
models to the behavior of the plant being simulated, the user will
need to adjust various parameters in the models. This list can be
useful in guiding calibration.
Aeration
Enter the hydraulic
head and headloss for the system, if known. Otherwise, use the
defaults. Calibrate with the blower efficiency and diffuser
headloss.
Pumping
Enter the hydraulic
head and headloss for the system, if known. Otherwise, use the
defaults. Calibrate with the pump efficiency and/or piping
headloss
Chemical Dosage &
Sludge Disposal
As these
calculations are a flat “flow multiplied by price” calculation, the
model is calibrated by adjusting the price parameter (flow is
calculated or set elsewhere).
Mixing & Heating
Energy
As these
calculations are done on a “per unit volume” basis, make sure that
you have the correct volume for the unit. Adjust the power
per unit volume usage as needed.
Miscellaneous Energy
The miscellaneous
energy uses are flat rates, so they are calibrated by adjusting the
rates as appropriate.
CHAPTER
14
This chapter describes the GPS-X optimizer and its associated
forms. More detail on the maximum likelihood objective function and
the statistical tests provided in the solution report can be found
in Appendix A and Appendix B, and the Optimizer
Solution Report. The complete list of nomenclature used
is found in Appendix C: Nomenclature. The references
cited in this chapter are listed in Appendix D:
References.
Optimizer Description
The optimizer is a module designed to minimize the value of a
user-selected objective function by adjusting the free variables in
this function. In the case of parameter estimation, these free
variables are the unknown process parameters. The optimizer uses
the Nelder-Mead simplex
method (Press et al., 1986) for minimization. The algorithm has
been modified to handle bounds on the optimization (i.e. free)
variables.
The simplex method is a
multi-dimensional procedure that does not rely on gradient
information. The algorithm searches through the multidimensional
"surface" using a direct search method to find a local minimum of
the objective function.
The procedure starts with
an initial point in the multi-dimensional parameter space and then
generates new points in space by perturbing the initial point a
scaled amount along each parameter direction. This leads to
p + 1 points in space that define a polyhedron (the simplex)
where p is the number of optimization variables. The points
are called the vertices of the simplex.
At each iteration, the simplex method reflects the vertex with the
highest function value (worst point) through the centroid of the
remaining p points of the polyhedron. The amount of
reflection is controlled by a reflection constant. If the
reflected vertex is the new best point (lowest function value) then
the polyhedron is expanded along the direction of reflection. The
amount of expansion is controlled by an expansion constant.
If the expanded vertex is better than the reflected vertex it is
taken as the new best point.
If after the reflection step the reflected vertex is worse than the
second worst vertex on the previous iteration, the polyhedron is
contracted. The reflected vertex is contracted through the centroid
of the remaining p vertices if it is the new worst point. If
the reflected vertex is the new second worst point, the worst point
is contracted through the centroid. The amount of contraction is
controlled by a contraction constant.
When a contraction step is
unsuccessful, the polyhedron is shrunk by moving the vertices
toward the best point. The amount of shrinkage is controlled by a
shrink constant.
At each step checks are
made to ensure that the parameter bounds have not been violated. If
they have been, the new vertex is moved back inside the bounds.
When the optimizer
termination criteria are satisfied, the optimizer runs one
additional simulation using the parameter values from the best
point found during the iteration process. There is the possibility
of encountering a local minimum that is not the global minimum of
the objective function. This is a problem that is common to most
optimization algorithms. As a result, it is important to solve
optimization problems (e.g. parameter estimation or process
optimization problems) from a number of different starting guesses
to ensure that the optimizer has found the global minimum.
If the user requests the
printing of confidence limits or derivative information (See
Summary of the Optimizer Settings and Parameters
section below) the optimizer conducts additional simulations to
generate numerical derivatives.
The constants that control the reflection, expansion, contraction,
and shrinkage of the polyhedron in the simplex method can be
accessed by selecting
Layout > General Data > System > Input Parameters
> Simulation Tool Settings and then clicking on the
More... button in the Optimizer sub-section. The
resulting form is shown in Figure 14‑1. You can
change any of the constant values if you wish. The parameter
entitled scaled step size in initial guess is used to
control the initial perturbation size.
Figure 14‑1
– Form Containing the Simplex Method Constants
The list of
objective functions available in GPS-X is given below:
Absolute
Difference
Equation 14.1
Relative
Difference
Equation 14.2
Sum of
Squares
Equation 14.3
Relative Sum of
Squares
Equation 14.4
Maximum
Likelihood
Equation 14.5
In the objective
function expressions given above the following nomenclature is
used:
zi,j
= the measured value of response j in experiment
i.
fi,j
= the value of response variable j predicted by the
process model in experiment i
gj
= the heteroscedasticity parameter for response j
m
= the number of measured response variables
nj
= the number of experiments (i.e. observations) for response
j
These objective function types are
accessed by clicking on the inverted triangle beside the
Optimize icon and then selecting Type from the
drop-down menu.
Using the Optimizer for
Parameter Estimation
In general, the maximum likelihood objective function should be
used when doing parameter estimation. This objective function
calculates statistically optimal parameter estimates based on
assumptions on the nature of the measurement errors. The sum of
squares objective function is a special case of the maximum
likelihood objective function derived using further simplifying
assumptions and can also be used. It is equivalent to the maximum
likelihood objective function when there is only one response or
target variable. Further details on parameter estimation and
the maximum likelihood and sum of squares objective functions can
be found in Appendix A of this chapter.
The other objective functions can
be used for curve fitting when calculating statistically optimal
parameter estimates is not a concern.
Using the Optimizer for
Process Optimization
In addition to data fitting applications, GPS-X can be used for
process optimization. For example, GPS-X can calculate the
operating conditions for your process model that optimize some
measure of process performance, such as operating cost or effluent
quality.
To solve this type of
problem in GPS-X you need to treat the problem as a data fitting
exercise. For example, if you want to minimize the value of a
certain model variable you need to specify an arbitrarily small
target value for the
model variable in a .dat file and have the optimizer
minimize the difference between the calculated variable value and
the target value using the absolute difference objective function.
This is equivalent to minimizing the model variable directly. The
target value should be made small enough so that the optimizer
cannot reach it.
You are not limited to using one
performance measure. You can select a number of different
performance measures and have GPS-X optimize these variables
simultaneously by fitting them to user-supplied targets.
There are four
different criteria used to terminate the optimizer:
1.
Parameter Tolerance: The maximum size along a
parameter dimension is defined as the difference between the
largest value of this parameter from among all of the simplex
vertices, and the smallest value of this parameter from among all
of the simplex vertices. This maximum size is scaled by dividing it
by the difference between the upper and lower bounds for this
parameter. If the maximum sizes for all of the parameter dimensions
are less than the parameter tolerance, the optimization process is
terminated.
2.
Objective function tolerance: If the range (largest minus
smallest) of objective function values covered by the simplex
vertices is less than the objective function tolerance, the
optimization process is terminated. Meeting this criterion without
satisfying criterion 1 may indicate that the objective function is
not very sensitive to some parameters.
3.
Scaled termination value for objective function: If the
objective function value at the current best point in the simplex
is less than the user-specified final objective function value, the
optimization process is terminated. This final value should be
positive and small in magnitude. This criterion is not used with
the maximum likelihood objective function. It is only applicable
for the other objective functions because their values have a lower
bound of zero (a perfect fit).
4.
Maximum number of optimizer iterations: The optimization
process is terminated if the maximum number of iterations is
reached.
Satisfaction of any
one of these four criteria will result in termination of the
iterative optimization process. These criteria provide a trade-off
between the length of the optimization run and the proximity to an
optimal solution.
The specific values used for
the termination criteria can be accessed from the
General
Data menu. Right-click in an empty region of the
drawing board to display the menu and select
System > Input Parameters > Simulation Tool
Settings to display the form shown in
Figure
14‑2.
Figure 14‑2
– Optimizer Form Containing the Termination Criteria Settings
A termination
criterion can be disabled by specifying a large negative value for
it. The objective function tolerance is disabled by
default.
The optimizer
module is equipped to handle three different types of process
measurements: time series measurements, long term operational data
that are averages of the original process measurements, and on-line
measurements. Each type of measurement set leads to a different
type of optimization problem in GPS-X. These optimization types can
be accessed by clicking on the inverted triangle to the right of
the Optimize and selecting Type.
Time Series
Optimization
This optimization type is the one normally used for both parameter
estimation and process optimization in GPS-X. It is designed to
handle both time series and steady-state measurements.
For this type of
optimization, you enter your measured data into a text file with a
.dat extension. This
text file should follow the naming and formatting conventions
discussed in the GPS-X User's Guide.
For parameter estimation
involving a dynamic model, the data entered into the text file will
be a set of time series values for each of the response variables.
In GPS-X the response variables are referred to as target
variables.
Steady-state optimization
is a time series-type optimization with only one data point
for each target variable. The steady-state solver is used and the
simulation has a stop time of 0.0. This type of optimization is
useful for calibrating the model to data reported as daily, weekly
or monthly averages. Data of this type are typically obtained from
composite samples and thus do not accurately reflect the time
dynamics of the real process. In a steady-state optimization, the
average data are used as the targets and selected model parameters
are adjusted to fit these targets.
When doing process
optimization, you enter single target values for your process
performance measures at the desired points in time. As mentioned
earlier, you should also make sure to use the absolute
difference objective function.
GPS-X will fit your model
to the measured data using the objective function that you select.
If you prepare output graphs to display the predicted values of the
model, GPS-X will automatically display the measured values
provided in the .dat file on the graphs.
GPS-X will draw a new
curve for the predicted values at each optimization iteration, in
order to track the progress of the optimizer. At the end of the
optimization process, final predicted responses are displayed so
that you can visually assess the fit. An example of this type of
graph is shown in Figure 14‑3.
Figure 14‑3
- Example GPS-X Output Graph Showing Measured Data (+ Markers) and
the Predicted Response (Continuous Line)
Dynamic Parameter
Estimation
GPS-X also has a sophisticated dynamic parameter estimation
procedure (DPE). DPE is designed for the estimation of time-varying
parameters. It can be used with on-line data or on a set of
off-line time series data. For details on using on-line data, see
the Advanced Control Module Manual.
The motivation behind DPE is that parameters in process models are
often not constant, but vary with time. For example, the oxygen
mass transfer coefficient in an aerated tank is often slowly
time-varying.
Dynamic parameter estimation is also useful for estimating
parameters in poorly understood processes. In these cases, the
model structure is likely to be incorrect. As a result, the model
may only be able to represent the data well over short time
intervals. In this case, using DPE will help compensate for the
model error and allow acceptable fitting of the measured data.
Another situation in which dynamic parameter estimation is useful
is when you are interested in detecting process changes and upsets.
If for example a model parameter is found to be relatively constant
during normal process operation but is sensitive to process
changes, you can track this parameter using the DPE feature and
on-line data to help provide an early warning of process changes or
disturbances.
In GPS-X, dynamic parameter estimation is done by applying the time
series optimization approach mentioned earlier to a moving time
window. Instead of estimating parameters from an entire set of
data, GPS-X calculates a set of parameter estimates for each time
window using the parameter estimates from the previous time window
as a starting guess. This approach can be used on a data file that
is continually updated with new blocks of data or on a static file
of time series data. You can use any of the objective functions
that are available for time series optimization when doing dynamic
parameter estimation.
The length of the time
window controls how often the parameters are updated. The shorter
the time window, the more often the parameters are updated.
When using short time windows, it may be necessary to filter the
data to eliminate noise so GPS-X does not fit the noise.
To ensure proper
termination of the optimization routine when using the DPE feature,
it is suggested that the time window and the communication interval
be chosen such that the time window is an integer multiple of the
communication interval.
This section
discusses the optimizer form that is accessed by selecting
General Data > System > Input Parameters >
Simulation Tool Settings > Optimizer > More…. The
upper portion of this form is shown in Figure 14‑2.
The lower portion of the form is shown below in Figure
14‑4.
Figure 14‑4
- Bottom of Optimizer Form
The termination
criteria found on this form were discussed under Termination
Criteria in this chapter. The form accessed from the
More... button in the Optimizer sub-section contains
the simplex method constants. The remaining settings and parameters
found in the Optimizer form are described below. You can
access the different sub-sections by scrolling down the form.
Optimizer Sub-Section
Number of optimized parameters: This sets the number of
variables to be adjusted by the optimizer. It should correspond to
the number of parameters specified as Optimize variables in
the Controls Setup window.
Number of data points
(at least 2): This number sets an upper bound on the number of
rows that GPS-X will read in from the .dat file. The default value is
large so that GPS-X can handle the majority of data sets without
having to change this value.
Detailed statistical
report (ON - OFF): If this option is set to ON and you
are using either the maximum likelihood or sum of squares objective
functions, the following statistics are provided in the solution
report in addition to the parameter estimates and objective
function value: variance-covariance matrix, correlation matrix, %
variation explained by regression, significance of the regression,
lack of fit test, observed values, predicted values, % error
between predicted and observed values, residuals, weighted
residuals, standardized residuals, standardized residual plots, and
the Portmanteau test on the weighted residuals. The
variance-covariance matrix and the correlation matrix are only
reported if the printing of confidence limits option is set
to ON. The lack of fit test is only reported if it is
turned ON. The Portmanteau test is only reported if it
is turned ON and the maximum likelihood objective function
is used. For more information on the statistical tests consult
Appendix B: The Optimizer Solution Report.
Solution report to
file (ON - OFF): If this option is set to ON, the
statistical output is appended to file stats.txt in the current
directory.
Maximum Likelihood
Sub-Section
Error distribution (Normal/Cauchy): Selects the
probability distribution of the measurement errors used in the
maximum likelihood objective function. The Normal option
should be used in most cases. The cauchy distribution looks similar
to the normal distribution but has heavier tails so that values far
removed from the mean have a higher probability than with they
would with a corresponding normal distribution.
Estimate standard
deviations of errors (ON - OFF): Determines if the
standard deviations should be estimated from the data using
Equation 14.10, or if they should be taken as
specified.
Standard deviation of
errors: Vector of standard deviations of errors associated with
each response variable (errors are assumed independent across
variables and observations). These values are not used in the
estimation problem if estimate standard deviations of
errorsis set to ON.
Use specified standard deviations as reference (ON -
OFF): Determines if the standard deviation of
errorsshould be used as reference values for the purpose of
counting the proportion of weighted residuals falling outside
reference bounds at a given level of significance. These reference
bounds are calculated using the reference standard deviations. In
addition, if this option is ONa chi-square test is performed
on the sum of squares of the standard deviations of the weighted
residuals divided by the reference standard deviations. This option
is set to OFFby default. It does not affect the outcome of
an optimization run.
Level of
significance: This value is used for computing the reference
bounds if use specified standard deviations as reference is
set to ON. This value cannot affect the outcome of an
optimization run.
Heteroscedasticity
model (ON - OFF): Determines if the variance model given
in Equation 14.8 is used. If this option is turned
off the heteroscedasticity parameters are set to zero.
Heteroscedasticity
parameters: A vector containing the heteroscedasticity
parameter for each response variable. If heteroscedasticity
model is set to OFF, all heteroscedasticity parameters
are ignored.
Derivative Information
Sub-Section
Report objective function gradient and Hessian(ON -
OFF): Controls the printing of the gradient and the Hessian of
the objective function at the solution. When this option is
ON, the gradient and Hessian are printed to the Log
window and if necessary to the stats.txt file. The gradient of the
objective function is a vector containing the derivatives of the
objective function with respect to each optimized parameter. The
relative gradient is a scaled version of the gradient. The
Hessian is a matrix containing the second derivatives of the
objective function with respect to the optimized parameters. A
Gauss-Newton Hessian approximation is used. The Hessian is only
reported when the maximum likelihood or the sum of squares
objective functions are used.
Report model sensitivity coefficients (ON - OFF):
Controls the printing of the model sensitivity coefficients at the
solution. When this option is ON the sensitivity
coefficients are printed to the Log window and if necessary
to the stats.txt file. For each target variable, the model
sensitivity coefficients are the derivatives of the target variable
with respect to each optimized parameter at each data point. These
derivatives are used in the calculation of the Hessian,
variance-covariance, and correlation matrices and the confidence
limits.
Finite-difference
relative perturbation size: The gradient vector elements and
the model sensitivity coefficients are calculated using a
finite-difference formula, specifically a forward-difference
formula. The step size used in calculating the derivatives is
calculated by multiplying the finite-difference relative
perturbation size by the absolute value of the parameter of
interest. Note that this setting affects the
calculation of the confidence limits.
Confidence Limits
Sub-Section
Printing of confidence limits (ON - OFF): Controls
the calculation and printing of confidence limits for the parameter
estimates. When this option is ON the confidence limits are
reported to the Log window and if requested to the stats.txt
file. When this option is OFF (default), the confidence
limits are not calculated or printed to the Log window or
the stats.txt file. Confidence limits are only reported when the
maximum likelihood or the sum of squares objective functions are
used. This switch does not affect the outcome of an optimization
run.
Confidence level for
confidence limits: The level of confidence used when
calculating the confidence limits. The default value of 0.95
corresponds to 95 percent confidence limits.
Treat the different
target variables as one target (ON - OFF): If there is
more than one target variable and the maximum likelihood objective
function is used, this switch controls whether the different
targets are treated as the same target for the purposes of
calculating the degrees of freedom used in the Student’s
t-statistic in the confidence limits calculations. This feature is
useful if you have a number of different target variables that are
actually the same variable measured in different experiments.
Significance of the
Regression Sub-Section
Level of
significance for significance of regression test: This
is the level of significance used in the significance of the
regression test. An appropriate message is printed by GPS-X
depending on the calculated probability value and the level of
significance. If the probability is larger than the chosen
significance level (default value is 0.05) it provides evidence
that the parameters are all zero and that the regression is not
significant for the corresponding target variable. If the
probability value is smaller than the significance level it
indicates that the regression is significant for the corresponding
target variable and that the variation explained by the regression
is greater than expected by chance. This test is only reported when
the maximum likelihood or the sum of squares objective functions
are used. It does not affect the outcome of an optimization
run.
Lack of fit
test (ON - OFF): Controls the printing of the lack of
fit test. If this option is set to ON and the detailed
statistical report is set to ON, the lack of fit test is
printed to the Log window and if necessary the stats.txt
file. This test determines whether the variance of the residuals is
acceptable compared to the user supplied estimate of the
measurement variance or the measurement variance calculated using
replicate measurements. This test is only reported when the maximum
likelihood or the sum of squares objective functions are used. It
does not affect the outcome of an optimization run.
Level of significance for lack of fit test: This is
the level of significance used in the lack of fit test. An
appropriate message is printed by GPS-X depending on the calculated
probability value and the level of significance. If the probability
for a certain target variable is smaller than the chosen
significance level, it indicates that there is a lack of fit
associated with this target variable.
Replication sum of
squares (User Supplied / Calculated): This option
allows the user to specify whether the replication sums of squares
used in the lack of fit test are user supplied or calculated by
GPS-X using repeat measurements.
Relative tolerance used
to detect repeat measurements: This is the relative difference
used when detecting repeat measurements in the data set. The
default value is 1.0E-4. Repeat measurements are inserted into
GPS-X by placing measurements very close together in time in the
.dat file. GPS-X
identifies repeat measurements by checking the relative difference
between all of the times entered into the .dat file. If the relative
difference between two time values is less than the replication
tolerance, the corresponding measurements are considered to be
repeats for the purposes of calculating the lack of fit test.
User-Supplied Replication
Sum of Squares Sub-Section
This sub-section is accessed by clicking on the More...
button in the Lack of Fit sub‑section.
Number of target
variables: This is the number of target variables used in the
optimization. It is only used to size the replication sum of
squares array and the degrees of freedom for replication sum
of squares array.
Replication sum of
squares: This is an array for entering a replication sum
of squares for each target variable. This is useful if you do not
have repeat measurements but you have a good estimate of the
measurement variance from past experience.
Degrees of freedom for
replication sum of squares: This is an array for
entering the degrees of freedom associated with each user supplied
replication sum of squares. If you do not have this information you
can enter rough estimates. Keep in mind that a large degrees of
freedom value implies a high degree of confidence in the variance
estimate.
Portmanteau
Sub-Section
Portmanteau test on weighted residuals (ON - OFF):
This switch turns the Portmanteau test on or off. The Portmanteau
test is used to detect trends in the weighted residuals. The
Portmanteau test is designed for data taken in sequence (e.g. time
or space). If trends are present, the residuals are not
independent. This violates one of the assumptions of the maximum
likelihood method and indicates that the model does not account for
all of the non-random variability in the data. An appropriate
message is printed depending upon the outcome of the test. This
test is only reported when the maximum likelihood objective
function is used. It does not affect the outcome of an
optimization run.
Maximum number of lags
used in portmanteau statistic: By default, the Portmanteau
statistic involves autocorrelations at lags up to half the length
of the time series. This setting can be used to impose further
restriction on the number of autocorrelations taken into account. A
large value effectively disables this option. This setting does not
affect the outcome of an optimization run.
Dynamic Sub-Section
DPE
Timewindow: This is the time window for dynamic parameter
estimation.
Before using the optimizer tool to estimate optimal parameter
values or optimize operating conditions, it is usually best to
experiment with manual
adjustment of the optimization variables. By conducting interactive
simulations you can observe the effects of the model parameters on
the response variables of interest. With this information you will
be able to make better judgments on appropriate variables to use in
a parameter estimation or optimization run.
For example, you can set
up an interactive simulation with slider controls for the
parameters and then try adjusting these variables to try and
achieve a visually acceptable fit. You can plot actual data along
with the target variables so that you can compare simulation and
actual data. This approach is useful for generating starting
guesses for parameter estimation and process optimization runs.
Introduction
In this appendix,
the maximum likelihood method is discussed in more detail. First a
general introduction on parameter estimation is given based on
material found in Bard (1974). This is followed by a detailed
discussion of the maximum likelihood method as implemented in
GPS-X. Finally, a brief description of the sum of squares objective
function is given.
Parameter Estimation
Parameter estimation is the procedure of fitting a mathematical
model of a process to measured data by calculating optimal
estimates of the model parameters. Parameter estimation differs
from simple curve fitting in that the criterion used to judge the
best fit is not arbitrary but is based on statistical
considerations. In addition, the model structure is based on
theoretical principles and the model parameters often have physical
significance. In curve fitting, the choice of model structure is
more arbitrary and is often chosen to simplify the computations.
The aim of parameter estimation is to not only fit a model to data
but to calculate parameter values that are good estimates of the
true values of the physical quantities.
Parameter estimation techniques can be applied to empirical models,
but the statistical properties of the estimates may not be as
meaningful in a physical sense. In addition, empirical models are
not as well suited for extrapolation as mechanistic models.
Parameter estimation is an
important step in the development of mathematical process models.
Process models contain parameters with physical significance that
may vary significantly from plant to plant. To develop process
models that can be used for predictive purposes, it is important to
estimate the unknown process parameters using measured process
data. Using literature values for the unknown parameters will often
result in a model that is not very useful for predicting actual
plant behavior.
Maximum Likelihood
Method
GPS-X uses the
maximum likelihood method for parameter estimation. In the maximum
likelihood method, the optimal parameter estimates are obtained by
maximizing the joint probability density function of the
measurements. This joint probability density function is a function
of the parameters and is known as the likelihood function. The form
of the likelihood function depends on the structure of the
experimental error.
In GPS-X it is assumed that the experimental errors are normally
distributed random variables with a mean of zero. As a result, the
likelihood function used
in GPS-X is:
Equation 14.6
where:
n =
number of experiments (i.e. observations)
m = number
of measured response variables
Vi
= variance-covariance matrix for the ith experiment
| Vi | = represents
the determinant of Vi
ei
= m x 1 residual vector that contains the differences
between the measured values of the response variables and the
values predicted by our mathematical process model.
q
= the vector of
parameters to be estimated in our mathematical process model.
This expression is derived
by substituting the residual vector, ei for
the error vector in the multivariate normal probability density
function (pdf). The error vector in the multivariate normal
pdf contains the differences between the measured values of
the response variables (the variables that we are fitting) and the
true values. See Bard (1974) for details.
GPS-X also assumes that the measurement errors are independent from
observation to observation and from response variable to response
variable. This means that the variance-covariance matrices found in
Equation 14.6 are diagonal but not necessarily
equal.
In GPS-X the log-likelihood function is used instead of the
likelihood function for mathematical convenience. Maximizing the
log-likelihood function is the same as maximizing the likelihood
function. The log-likelihood function is derived by taking the
natural logarithm of Equation 14.6, as shown
below:
Equation 14.7
where:
= variance of response j in
experiment i
In this derivation, it is assumed that the variance of a response
variable is not constant across all the observations. To take this into account, GPS-X
uses the following expression for the estimated variance (Reilly
et al., 1977):
Equation 14.8
where:
= estimate of the variance,
wj
= proportionality constant that will be called the standard
deviation of the weighted residuals for response j
fi,j
= value of response variable j predicted by the process
model in experiment i
gj
= heteroscedasticity parameter for response j
This expression relates the variability of response variable
j to the magnitude of the predicted value for response
j.
The heteroscedasticity parameter controls how the variance depends
on the predicted values. This parameter is bounded between 0 and 2
and is continuous within this range. A value of 0 indicates
constant absolute variability across all of the observations for
response variable j. A value of 2 indicates constant
relative variability across all of the observations for response
variable j.
To determine the optimal value of wj,
given the other adjustable parameters in the log-likelihood
function, the estimated variance given by Equation
14.8 is substituted into Equation 14.7.
The resulting equation is then differentiated with respect to
wj
and the derivative is set to zero, leading to the following
expression:
Equation 14.9
where:
= measured value of response
j in experiment i
Substituting this expression into Equation 14.8
yields the estimate for the variance that accounts for
non-homogeneous measurement errors:
Equation 14.10
If Equation 14.10 is substituted into Equation
14.7, the log-likelihood function becomes:
Equation 14.11
To allow for the possibility that the number of observations is
different for each response variable, Equation 14.11
is re-written as (Steiner et al., 1990):
Equation 14.12
This is the function that is maximized by GPS-X to fit process
models to measured data and obtain optimal parameter estimates.
Equation 14.12 is a measure of the probability that
the measurements were generated by the process model.
Note that the user
can set the heteroscedasticity factors or have GPS-X estimate their
optimal values for you.
As process models are
generally nonlinear, the parameter values that maximize
Equation 14.12 cannot be determined analytically. An
iterative optimization method is required to maximize
Equation 14.12. As mentioned earlier in this chapter,
GPS-X uses the Nelder
and Mead simplex method (Press et al., 1986) for
optimization. This method is a type of direct search algorithm that
does not require the calculation of partial derivatives. This is
helpful when estimating parameters in systems of differential
equations. The simplex method also has the advantage that it can
handle objective functions containing discontinuities. This method
is generally slower than derivative-based optimization methods, but
can handle a greater variety of objective functions and is often
found to be quite robust in finding solutions. The version of the
simplex method implemented in GPS-X allows for bounds to be placed
on the parameters.
Since the simplex method is designed for minimization, GPS-X
minimizes the negative of Equation 14.12 to determine
the optimal parameter estimates.
Sum of Squares Objective
Function
When there is only
one target variable and the variance is constant across all
observations (i.e. the heteroscedasticity parameter is zero), the
sum of squares objective function is equivalent to the maximum
likelihood objective function. For problems with more than
one target variable, the sum of squares objective function is a
special case of the maximum likelihood objective function that
results if we make additional assumptions about the measurement
errors. Our maximum likelihood objective function, given by
Equation 14.12, is derived using the following
assumptions:
·
The measurement errors are normally distributed random variables
with a mean of zero.
·
For each target variable, the measurement errors are independent
from observation to observation. Each target variable has its
own variance which varies from observation to observation according
to a power-law. The variances are unknown and are calculated
as part of the optimization process.
·
There is no correlation between different target variables
If we make the additional assumptions that all of the variances are
equal (i.e. all responses have same variance) and the variances do
not change from observation to observation (i.e. the
heteroscedasticities are all zero) then the maximum likelihood
function reduces to the sum of squares objective function in the
multi-response case. The assumptions used to derive the sum
of squares objective function in the multi-response case do not
apply in most practical situations. Therefore, it is
recommended that the maximum likelihood objective function be used
for calibration problems with more than one target
variable.
Introduction
In this appendix,
the solution report provided in the Log window after a GPS-X
optimization run is discussed. The solution report provides the
user with the solution found by the optimizer and a number of
additional statistics that are valuable when doing parameter
estimation. This report also may be printed to the stats.txt file. First we discuss
the basic report and the detailed statistical report. This is
followed by a summary on using the statistical tests and a
discussion of over parameterization.
NOTE:
In the solution report, the parameters are given generic names and
are numbered according to the order that the parameters are listed
in the Control window containing these parameters. For
example, the parameter at the top of the Control window is
labeled Parameter 1.
Many of the
statistics in the solution report are organized according to target
variable names. Generic target variable names are used and the
names are numbered according to the order given for the target
variables in the target variables form that is accessed by
selecting Target Variables... from the Optimize
drop-down menu. For example, statistics corresponding to the first
target variable in the form are presented under the heading
Target 1.
Basic Report
The basic solution
report, which is printed to the Log window, includes the
number of iterations required by the optimizer, the initial and
final values of the objective function, the initial and final
values of the parameters (i.e. the optimization variables), the
heteroscedasticity parameter values, and a summary of the chosen
optimizer settings. In addition, the statistics discussed below are
provided if requested in the Optimizer form.
Gradient of the Objective
Function
The gradient vector
of the objective function is defined as:
Equation 14.13
where:
F
= the objective function
θ1,
2 /
θp
= the optimization variables
In the case of
parameter estimation these variables are the parameters being
estimated. The vector contains
the optimization variables.
The partial
derivatives are calculated numerically using the following
forward-difference formula:
Equation 14.14
where:
k =
1
hk
= the step or perturbation size
The step size is
calculated using the following formula:
Equation 14.15
In Equation 14.15, the value of 10-7 is
referred to as the finite-difference relative perturbation
size in GPS-X and can be changed in the Derivative
Information sub-section of the Optimizer form.
The gradient provided in the solution report is calculated at the
solution. If the solution is a local minimum, the elements of
the gradient vector should be close to zero. Because poor
scaling of the parameters and the objective function can make the
gradient appear large, the relative gradient is also
reported. The elements of the relative gradient are
calculated by scaling the gradient elements using the following
equation:
Equation 14.16
The infinity norm of the relative gradient vector is also
reported. If the solution is a local minimum, this norm
should be close to zero.
The gradient is
only reported when the report objective function gradient and
Hessian switch in the Derivative Information sub-section
in the Optimizer form is set to ON.
Model Sensitivity
Coefficients
The model
sensitivity coefficients express the local sensitivity of the
process model to infinitesimal changes in the optimization
variables (subset of the model parameters). They are the partial
derivatives of the model with respect to the optimization
variables. A sensitivity coefficient can be calculated for each
target (i.e. response) variable with respect to each optimization
variable at each data point. The sensitivity coefficients are
calculated using the forward-difference approximation shown
below:
Equation 14.17
In this equation fi,j is the jth target
variable at the ith data point. The step size is calculated
in the same way as for the gradient.
The sensitivity coefficients are
only reported when the report model sensitivity
coefficientsswitch in the Derivative
Informationsub-section in the Optimizer form is set to
ON.
Hessian of the Objective
Function
The Hessian is a matrix of second partial derivatives of the
objective function with respect to the optimization variables. For
the case of two optimization variables, the Hessian is defined
as:
Equation 14.18
In GPS-X the elements of the Hessian are calculated using the
Gauss-Newton approximation. For the sum of squares objective
function, each Hessian element is defined as:
Equation 14.19
where:
m = number
of target or response variables
nj
= number of data points for target variable j
For the maximum likelihood objective function, Equation
14.5, each Hessian element is defined as:
Equation 14.20
where γj is the heteroscedasticity of the
jth response variable.
The variable ei,j is the residual and is defined
as:
Equation 14.21
The variable zi,j is the measured value of
response j at the ith data point.
The expressions in Equation 14.19 and Equation
14.20 were derived assuming that the response variables are
equivalent to the state variables. Even if this is not the case,
this is not a limitation because the sensitivity coefficients are
being calculated using finite-differences. Therefore any
dependencies in the responses are being taken into account.
The Hessian matrices for
the other objective functions are not calculated because they
cannot be approximated using the Gauss-Newton approximation. As a
result, they require the calculation of second order
sensitivities.
If any of the optimization
variables are at their bounds, the elements of the Hessian
involving these variables are zero.
The Hessian matrix is only reported when the report objective
function gradient and Hessianswitch in the Derivative
Informationsub-section in the Optimizer form is set to
ON.
Confidence Limits
GPS-X calculates linear-approximation confidence limits for the
parameter estimates using the variance-covariance matrix. For the
sum of squares objective function, the variance-covariance matrix
of the parameter estimates is defined as (Bard, 1974):
Equation 14.22
The matrix Ĥ-1 is the inverse of the Gauss-Newton Hessian approximation
to the sum of squares objective function at the solution. The
variable s2 is the variance estimate for the
measurement errors and is defined as:
Equation 14.23
where
is the value of the sum of squares objective
function at the solution, n is the number of data points
(total number considering all target variables), and p is
the number of parameters (i.e. optimization variables).
For the maximum likelihood objective function, the
variance-covariance matrix of the parameter estimates is defined as
(Bard, 1974):
Equation 14.24
The matrix Ĥ-1 is the inverse of the Gauss-Newton
Hessian approximation to the maximum likelihood objective function
at the solution.
For both the sum of squares and the maximum likelihood objective
functions, the 100(1-a) % confidence limits for parameter
k are:
Equation 14.25
The variable t(n – p; a / 2) is the Student’s
t-statistic, with n-p degrees of freedom, and a
significance level of a / 2. In this case, n is
the number of data points per response, except for the sum of
squares objective function where it is the total number of data
points. The variable a is the significance level for
the parameter estimates. It is defined by the following
expression:
Equation 14.26
where δ is the confidence coefficient. The confidence
coefficient corresponds to the confidence level for confidence
limits parameter found in the Confidence Limits
sub-section of the Optimizer form. The default value is 0.95
so that by default GPS-X reports 95 percent confidence
limits.
Equation 14.27
The confidence
limits are only reported when the printing of confidence
limitsswitch in the Confidence Limitssub-section in the
Optimizer form is set to ON.
Detailed Statistical
Report
The detailed
statistical report provides additional information that is helpful
when doing parameter estimation. It allows the user to assess how
well their model fits the measured data. The detailed statistical
report is activated by setting the detailed statistical
reportoption to ON in the Static sub-section of
the Optimizer form. The statistics included in the detailed
statistical report are discussed below.
Variance-Covariance and
Correlation Matrices
The variance-covariance matrix is defined by Equation
14.22 for the sum of squares objective function and by
Equation 14.24 for the maximum likelihood objective
function. This matrix contains the variances of the individual
parameter estimates (diagonal elements) and the covariances between
the different parameter estimates (off-diagonal elements).
Parameters with large variances may be unnecessary.
The elements of the
correlation matrix are calculated as follows:
Equation 14.28
where
is the (k, l) element of the
variance-covariance matrix.
By definition, the elements of the correlation matrix are always
between zero and one. Unless some parameters are on their bounds
the diagonal elements are always equal to one. The off-diagonal
elements indicate the degree of correlation between pairs of
parameters. A large off-diagonal element indicates a high degree of
correlation between a pair of parameters. This provides evidence
that some of the parameters in the model are unnecessary and the
model is over parameterized. This may indicate that the model is
not adequate for the task at hand or that the data do not provide
enough information to allow estimation of all of the model
parameters (Draper and Smith, 1981).
The
variance-covariance and correlation matrices are only reported when
the printing of confidence limitsswitch in the Confidence
Limitssub-section in the Optimizer form is set to
ON.
Percentage Variation
Explained by Regression
This is a statistic often reported by linear least squares
packages. In GPS-X it has been modified to handle nonlinear maximum
likelihood problems with heteroscedasticity. The statistic, adapted
from Steiner et al. (1990), is calculated as follows
for each target variable:
Equation 14.29
It is multiplied by 100 to get a percentage. The variable
SSj is the weighted residual sum of squares for
response j and is defined by:
Equation 14.30
The variable SSmeanj is the total weighted sum of
squares corrected for the mean and is defined below:
Equation 14.31
where is a weighted average of the measured values
for response j and is calculated as:
Equation 14.32
The overall variation explained is calculated as:
Equation 14.33
The % variation
explained is not a model adequacy test but is useful because it
shows how much of the variation in the data is accounted for by the
fitted model. This test is only reported when the maximum
likelihood or sum of squares objective functions are used.
Significance of
Regression
This test involves testing the null hypothesis that all the model
parameters are zero against the alternative hypothesis that all the
parameters are not zero.
The following sum of squares ratio
is calculated for each response variable:
Equation 14.34
where SSregj is the weighted regression sum of
squares for target variable j and is defined by:
Equation 14.35
The sum of squares
ratio follows an F-distribution if the errors in the
measurements are independent, normally distributed random variables
and if the model parameters are all zero.
GPS-X reports the F-value for each target variable and reports the
probability that it is from an F-distribution with p-1 and
degrees of freedom. If the probability is
larger than the chosen significance level it provides evidence that
the parameters are all zero and the regression is not significant
for the corresponding target variable. The significance level is
entered in the General Data > System > Parameters >
Optimizer form in the Significance of the Regression
sub-section.
If the probability value is smaller than the significance level it
indicates that the F-value does not follow an F-distribution and
therefore we accept the alternative hypothesis that the model
parameters are not all zero. In this case the regression is
significant for the corresponding target variable and the variation
explained by the regression is greater than expected by chance. For
a model to be useful for predictive purposes the F-values should be
large and the probability values should be close to zero.
GPS-X provides messages that summarize the results
of the significance test. The significance test is based on linear
regression theory and is only approximate for nonlinear models.
This test is only reported
when the maximum likelihood or the sum of squares objective
functions are used.
Lack of Fit Test
This test determines whether the variance of the residuals is
acceptable compared to the user supplied estimate of the
measurement variance or the measurement variance calculated using
replicate measurements. This serves as a model adequacy test.
For each target variable,
the following sum of squares ratio is calculated:
Equation 14.36
The variable SSrepj is the weighted replication
sum of squares for target variable j and is calculated
as:
Equation 14.37
Where n is the number of independent variable values for
which there are replicates, nrv is the number of
replicates for the vth value of the independent variable,
is the mean of the replicate
measurements corresponding to the vth variable,
zuy is the uth replicate measurement
corresponding to the vth variable, and fvj
is the predicted value of response variable j corresponding
to the vth value of the independent variable.
The variable repdofj is the degrees of freedom
for the weighted replication sum of squares for target variable
j. It is defined as:
Equation 14.38
As mentioned earlier, in this chapter, repeat measurements are
inserted into GPS-X by placing measurements very close together in
time in the .dat
file. See the previous Lack of Fit
Sub-Section, located in this chapter, for
details.
The weighted lack of fit
sum of squares for the jth target variable,
SSlofj, is calculated as:
Equation 14.39
As mentioned in
this chapter, the user also can provide the replication sum of
squares and the replication degrees of freedom in the lack of fit
sub-section of the Optimizer form. In order for these values
to be used by GPS-X, the replication sum of squares option
in the Lack of Fit sub-section of the Optimizer form
must be set to User Supplied.
The sum of squares ratio or F-value follows an F-distribution if
the errors in the measurements are independent, normally
distributed random variables and if the weighted lack of fit sum of
squares is not much larger than the weighted replication sum of
squares.
GPS-X reports the F-value for each target variable
and reports the probability that it is from an F-distribution with
nj – p ‑ repdofj and
repdofj degrees from freedom. If the
probability for a certain target variable is smaller than the
chosen significance level, it indicates that there is a lack of fit
associated with this target variable. The significance level
is entered in the General
Data>System>Parameters>Optimizer form in the
Lack of Fit sub-section.
GPS-X provides messages
that summarize the results of the lack of fit test. The lack of fit
test is based on linear regression theory and is only approximate
for nonlinear models.
This test is only
reported when the maximum likelihood or the sum of squares
objective functions are used and the lack of fit switch in
the Lack of Fit sub-section in the Optimizer form is
set to ON.
Observed and Predicted
Values
The observed values
are the measured target variable values provided in the
.dat file. The
predicted values are the target variable values calculated by the
model that correspond to the observed values (i.e. they are
calculated at the same points in time).
% Error between Predicted
and Observed Values
The % errors are useful for detecting any large discrepancies
between the model and the measured data. The % error statistic is
calculated as follows:
Equation 14.40
where %Ei,
j is the percent error between zi, f and
fi,j
Residuals
As discussed
earlier, the residuals are defined as:
Equation 14.21
They are an
estimate of the error in the measurement, assuming that the model
is correct.
The weighted residuals and standardized residuals are only reported
when the maximum likelihood objective function is used. The
weighted residuals are defined as:
Equation 14.41
These are scaled
residuals that are not as dependent on the magnitude of the
measurements and the predicted values as are the unscaled
residuals.
The standardized
residuals are defined as:
Equation 14.42
The variable Si,j is the estimated variance of
response j for the ith data point and is defined in
Equation 14.10.
The standardized residuals and their associated plots can be used
to check whether the residuals are independent and normally
distributed. This assumption is fundamental to the development of
the maximum likelihood objective function used in GPS-X. The
standardized residuals are normalized and should have a mean of
zero and a variance of one if the residuals are normally
distributed.
To check for violation of the independence assumption, the
standardized residual plots should be examined for any noticeable
trends. The presence of trends provides evidence of serial
correlation. Serial correlation occurs when residuals taken in
sequence are correlated with each other.
To check the assumption that the errors are normally distributed,
the standardized residual plots should be examined to see if the
residuals are randomly scattered about zero. In addition, the
values of the standardized residuals should be examined to check
whether approximately 95 percent of the residuals lie between +2
and -2.
The standardized residual
plots are scaled to fit in the Log window. The largest
standardized residual for each response variable is represented by
the maximum number of " * " characters, which is seven. The
remaining standardized residuals for a response variable are scaled
relative to this maximum residual for the purposes of calculating
how many " * " characters to print. You should consult the actual
values of the standardized residuals provided in the Log
window to determine whether the actual magnitudes of the residuals
for a response variable are large.
If the assumption that the residuals are independent and normally
distributed is violated it provides evidence that our model is
inadequate to represent the experimental data. If the process model
structure is correct, it should account for the non-random
variability in the data.
The assumption that the residuals should be independent if the
measurement errors are independent is not strictly correct but is
fine for practical purposes. The residuals are always correlated to
some extent as a result of the fact that there are n
measurements but only (n - p) degrees of freedom,
where p is the number of parameters to be estimated (Draper
and Smith, 1981).
Portmanteau Statistic
The Portmanteau
test is used to detect trends in the weighted residuals. If trends
are present, the residuals are not independent. This violates one
of the assumptions of the maximum likelihood method and indicates
that the model does not account for all of the non-random
variability in the data. The Portmanteau test is designed for data
taken in sequence (e.g. time or space).
The Portmanteau statistic for response variable j is
calculated as the number of observations times the sum of the
squared autocorrelations between the weighted residuals up to a
certain number of time lags (Brockwell and Davis, 1996):
Equation 14.43
where
nj is the number of data points for response
j, τ is the number of time lags, and
ρj(k) is the sample autocorrelation
among the weighted residuals up to lag k for
response j.
The sample autocorrelation is defined as:
Equation 14.44
where Ωj(k) is the sample auto covariance
function between the weighted residuals up to lag k for
response j and is defined as:
Equation 14.45
The sample
autocorrelations are approximately independent, normally
distributed random variables with mean zero and a variance of
1/nj if the weighted residuals are independent
and identically distributed (Brockwell and Davis, 1996).
This approximation gets better as the number of measurements
increases. Although the residuals themselves are not necessarily
identically distributed (they can have different variances), the
weighted residuals should be identically distributed because they
are scaled.
If the sample autocorrelations for a response variable are
independent and normally distributed as mentioned above, then the
variables should be
independent and normally distributed with mean zero and variance
one. Since the Portmanteau statistic is a sum of squares of these
variables, it is distributed as a chi-squared variable.
GPS-X reports the
Portmanteau statistic for each target variable and reports the
probability that it is from a chi-squared distribution with
degrees of freedom. If the probability for a certain target
variable is smaller than the chosen significance level it provides
evidence of a trend in this target variable's weighted residuals.
The significance level is entered in the General Data >
System > Parameters > Optimizer form in the
Portmanteausub-section.
GPS-X provides
messages that summarize the results of the Portmanteau test. This
test is only reported when the maximum likelihood or the sum of
squares objective functions are used and the portmanteau test on
weighted residualsswitch in the Portmanteausub-section
in the Optimizer form is set to ON.
Summary on Using the
Statistical Tests
See Table
4‑1 for a summary of how to use the statistics given in the
solution report to assess the adequacy of the fitted model. Keep in
mind that many of the tests become more reliable as the number of
measurements increases. Even if the tests indicate that the model
is not adequate, it does not mean that you cannot use the model. A
visual inspection of the plots provided by GPS-X, showing the
measured values and the predicted values, may indicate that the
model captures the major trends in the data. This is often good
enough for practical purposes as it may only be important to model
certain aspects of a physical system.
Overparameterization
Overparameterization occurs when there are more parameters in the
process model than necessary to fit the data. This situation leads
to correlations between model parameters as mentioned earlier in
the context of the correlation matrix. As a result, the objective
function near the solution to the parameter estimation problem has
elongated contours and the solution is not very sensitive to
changes in certain parameters.
The user should be careful when
choosing the adjustable parameters in a parameter estimation run.
The model should be sensitive to these parameters. It is often not
practical to select the entire model parameters as optimization
variables because this slows the optimization process, and will
likely result in meaningless values for certain model parameters.
It is preferable to choose only those parameters that have the
greatest affect on the mismatch between the model and the data. See
the Sensitivity Analysis chapter in the GPS-X
User's Guide for details on how to conduct a sensitivity
analysis of your model using GPS-X before doing a parameter
estimation run.
Table 14‑1 - Summary of How to Use the
Statistical Tests
Statistic
|
How to Interpret
|
Confidence Limits
|
If the
confidence limits on a parameter include zero it indicates that
this parameter is not significant to the model.
|
Correlation Matrix
|
Large
off-diagonal elements (close to one) suggest that certain
parameters are correlated.
|
Significance of Regression Test
|
Probability values larger than the significance level indicate that
the variation explained by the regression is less than expected by
chance.
|
Lack of Fit Test
|
Probability values smaller than the significance level provide
evidence for a lack of fit in the fitted model.
|
Standardized Residuals
|
The
majority of standardized residuals should be between +2 and -2.
|
Standardized Residual Plots
|
The
residuals should be randomly scattered about zero without any
noticeable trends.
|
Portmanteau Probability
Values
|
Probability values smaller than significance level suggest that the
residuals are correlated.
|
a
|
=
|
significance level for the
parameter estimates
|
|
=
|
correlation matrix at the
solution
|
d
|
=
|
confidence coefficient
|
ei
|
=
|
m x 1 residual vector that
contains the differences between the measured values of the
response variables and the values predicted by the mathematical
process model
|
|
=
|
|
%Ei,j
|
=
|
% error between the measured and
predicted values at point for response j in experiment
i
|
fi,j
|
=
|
value of response variable
j predicted by the process model in experiment i
|
|
=
|
objective function evaluated at
qk
|
|
=
|
the partial derivative of the
objective function with respect to parameter qk
|
|
=
|
partial
derivative of respect to fi,j with respect to
parameter qk
|
|
=
|
second partial derivative of the
objective function with respect to qk and
ql
|
|
=
|
the heteroscedasticity parameter
for response j
|
|
=
|
the step or perturbation size used
in the forward-difference derivative formula for the kth
parameter
|
H
|
=
|
Hessian matrix of objective
function
|
|
=
|
Hessian matrix of objective
function at the solution
|
m
|
=
|
the number of measured response
variables
|
|
=
|
the number of experiments (i.e.
observations) for response j
|
n
|
=
|
the number
of experiments (i.e. observations) assuming all responses have the
same number of observations
|
|
=
|
number replicates for the
vth value of the independent variable
|
h
|
=
|
number of independent variable
values for which there are replicates
|
|
=
|
standard deviation of the weighted
residuals for response j
|
|
=
|
sample autocovariance function
between the weighted residuals up to lag k for response
j
|
p
|
=
|
number of parameters being
estimated
|
|
=
|
Portmanteau statistic for response
j
|
|
=
|
sample autocorrelation among the
weighted residuals up to lag k for response j
|
|
=
|
% variation explained by
regression expressed as a fraction (subscript used if referring to
a response variable)
|
RelGradk
|
=
|
relative
gradient for parameter
|
repdof
|
=
|
weighted replication sum of
squares degrees of freedom (subscript used if referring to a
response variable)
|
|
=
|
the variance of response j
in experiment i
|
|
=
|
the estimate of the variance,
|
|
=
|
standard error of parameter
at the solution
|
SS
|
=
|
weighted residual sum of squares
(subscript used if referring to a response variable)
|
SSlof
|
=
|
weighted lack of fit sum of
squares (subscript used if referring to a response variable)
|
SSmean
|
=
|
total weighted sum of squares
corrected for the mean
|
SSreg
|
=
|
weighted regression sum of squares
(subscript used if referring to a response variable)
|
SSrep
|
=
|
weighted replication sum of
squares (subscript used if referring to a response variable)
|
t(a,b)
|
=
|
student's t statistic with
a degrees of freedom and a significance level of
b
|
t
|
=
|
number of time lags used in the
Portmanteau statistic
|
|
=
|
the vector of parameters to be
estimated in the process model
|
|
=
|
vector of estimated parameters at
the solution
|
|
|
variance-covariance matrix at the
solution
|
|
=
|
the variance-covariance matrix for
the ith experiment
|
|
=
|
the
determinant of
|
|
=
|
the measured value of response
j in experiment i
|
|
=
|
mean of the replicate measurements
corresponding to the vth value of the independent
variable
|
|
=
|
uth
replicate measurement corresponding to the vth value of the
independent variable
|
Bard, Y. A., Nonlinear Parameter
Estimation. Academic Press, New York (1974).
Brockwell, P. J. and Davis, R. A. Introduction to
Time Series and Forecasting. Springer-Verlag, New York
(1996).
Draper, N. R., and Smith, H. Applied Regression
Analysis. John Wiley & Sons, (1981).
Edgar, T. F. and Himmelblau, D. M. Optimization of
Chemical Processes. McGraw-Hill, (1988).
Press, W. H., Flannery, B. P., Teukolsky, S. A., and
Vetterling, W. T. Numerical Recipes: The Art of Scientific
Computing. Cambridge University Press, New York
(1986).
Reilly, P. M., Barjramovic, R., Blau, G. E., Branson,
D. R., and Saverhoff, M. J. Guidelines for the Optimal Design of
Experiments to Estimate Parameters in First Order Kinetic Models.
Can. J. Chem. Eng., 55, 614 (1977).
Steiner, E. C., Rey, T. D., and McCroskey, P. S.
SimuSolv Reference Guide. The Dow Chemical Company, Midland, MI,
Vol. 2 (1990).
Dochain D. and Vanrolleghem P.(2001)
Dynamic Modelling and estimation in wastewater treatment Processes,
IWA Publishing, London, UK
Hauduc H, Neumann M. B., Muschalla D.,
Gamerith V., Gillot S. and Vanrolleghem P. (2011), Towards
quantitative quality criteria to evaluate simulation results in
wastewater treatment- A critical review, 8th IWA
symposium on systems analysis and integrated assessment,
Watermatex, pp 37-49
From MBR supplementary:
Membrane-Coupled Activated Sludge
System: The Effect of Floc Structure on Membrane Fouling.
Separation Science and Technology, 34(9),
pp.1743-1758.
Choi, S., Yoon, J., Haam, S., Jung,
J., Kim, J., Kim, W. (2000). Modelling of the Permeate Flux
during Microfiltration of BSA-Adsorbed Microspheres in a Stirred
Cell. Journal of Colloid and Interface Science, 228,
pp. 270-278.
Garcia, G.E., Kanj, J.
(2002). Two Years of Membrane Bioreactor Plant Operation
Experience at the Vejas Tribe Reservation. Proceedings of
the Water Environment Federation’s 75th Annual Technical
Exhibition and Conference, September 28th – October 2, Chicago,
Illinois.
Günder, B. (2001).The
Membrane-Coupled Activated Sludge Process in Municipal Wastewater
Treatment. Technomic Publishing Company, Lancaster, PA,
USA.
Merlo, R.P., Adham, S., Gagliardo,
P., Trussell, R.S., Trussell, R. (2000). Application of
Membrane Bioreactor Technology for Water Reclamation.
Proceedings of the Water Environment Federation’s 73rd
Annual Technical Exhibition and Conference, October 14 - 18,
Anaheim, CA.
Shin, H-S., Lee, W-T., Kang, S-T.
(2002). Formation of Dynamic Membrane in Submerged Membrane
Bioreactors (MBRs). Proceedings of the Water Environment
Federation’s 75th Annual Technical Exhibition and Conference,
September 28th – October 2, Chicago, Illinois.
Wallis-Lage, C., Steichen, M.,
deBarbadillo, C., Hemken, B. (2005). Shopping for an MBR –
What’s for sale? Water Environment & Technology,
17(1), pp. 31-35.
CHAPTER
15
Introduction
The simulation results from dynamic wastewater treatment plant
models are usually compared to the measured time-series datasets
during model calibration and validation studies. Typically, the
comparison between the dynamic simulation results and the time
series measured data is based on qualitative assessment that
involves visual inspection. Although, visual inspection based
qualitative assessment is useful method to check the fit between
the measured data and simulation results, the methods does not help
in differentiating between goodness of one simulation results over
the others when many simulation results can pass the qualitative
assessment. In such situations it may be more meaningful to use
quantitative statistical criteria to select the results of one
calibration parameter set over the other.
Considering the
usefulness of quantitative statistical criteria in evaluating and
improving the model calibration and validation, a set of selected
statistical indices were implemented in the GPS-X v6.4. The direct
calculation of statistical indices inside GPS-X will significantly
reduce the efforts of exporting and manipulating the simulation and
measured data for estimation of the statistical indices. The
statistical measures implemented will be explained in following
sections using the predicted and measured BOD data as shown in a
typical GPS-X output graph (see Figure 15‑1).
Figure 15‑1
– Typical Time Series Plot of Predicted and Measured Data
The time-series
measured data is considered to contain, a time stamp at which the
sample was collected and a value of the data. The time-series data
could be available in a regular or irregular time intervals. The
data measured at a wastewater plant could be from different type of
sampling. The following data types are common:
1.
Grab Data: The sample is collected at a specific time.
The on-line data and data from grab samples belong to this
category. The data point is fully defined by the time stamp and the
measured value.
2.
Time proportional composite data: Equal volumes of
samples are collected at regular time intervals. These samples are
mixed to give a composite sample. The data point is completely
defined by the duration of composite and time interval for sample
collection.
3.
Flow proportional composite data: Equal volumes of
samples are collected at equal volumes of wastewater treated. This
means more number of samples is collected during high flow
conditions and lower number of samples during the low flow
conditions. These samples are mixed to give a composite sample. The
data point is completely defined by the duration of composite and
volume interval for sample collection.
As compared to the
three measured data types indicated above, the raw simulation data
produced by the wastewater models refers to a specific time and
hence is similar to the Grab data type. Therefore, when simulation
data is required to be compared with the composite measured data
types (Type 2 and 3 above), a further processing of simulation data
is necessary for meaningful comparison.
Considering the
above discussion, depending on the type of the sample used for
comparison, an equivalent sample data estimated using the
simulation data in the statistical analysis. The type of sample can
be specified in “Measured Data Type” section of the
statistical setup menu (Figure 15‑2).
Figure 15‑2
- Statistical Analysis Set up Menu for Data Type, Output Plots and
Table
Table of Statistical
Indices
The main objective
in this first implementation is to focus on the most important
statistical indices that may be helpful in quantitative assessment
of model calibration and validation. A brief description of
the statistical indices that are calculated for a given dataset is
provided in Table 15‑1.
Table 15‑1 –
Statistical Measures and Equations
No.
|
Statistical Measure
|
Equation
|
1
|
Mean of Residuals
|
|
2
|
Mean of Absolute
Residual
|
|
3
|
Mean of Squared Residual
|
|
4
|
Absolute Maximum Residual
|
|
5
|
Root of Mean Squared Residuals
|
|
6
|
Mean of Relative Residual
|
|
7
|
Mean of Absolute Relative
Residual
|
|
8
|
Mean of Squared Relative
Residual
|
|
9
|
Relative Volume Residuals
|
|
10
|
Absolute Relative Volume
Residuals
|
|
11
|
Theil's Inequality Coefficient
|
|
12
|
Nash-Sutcliffe (R2)
|
|
13
|
Standard deviation of Residuals
(SDR)
|
|
14
|
Mean of Standardized
Residuals
|
|
Legend:
Oi
= the observed (measured) value
Pi
= the predicted (simulated) value
Om =
mean of the observed (measured value)
n =
number of data points
i
= the ith observation
Ri
= (Oi - Pi) = residual (error)
MR = Mean of
Residuals
A typical GPS-X
output showing the calculated statistical indices given quality
indicator is as shown in Figure 15‑3.
Figure 15‑3 -
Summary of the Statistical Measures Calculated for a Time Series
Dataset
Predicted vs.
Measured
In this plot the
predicted values are plotted against the measured values. A 45
degree line is also placed on this graph to see the deviation of
data points from this line. Assuming a perfect fit, the plotted
data points will plot on the 45 degree line. Visual inspection of
this plot can highlight the model biases and other systematic
errors in the model. A typical output from GPS-X analysis is as
shown in Figure 15‑4.
Figure 15‑4
- Predicted vs. Measured Data Plot
For a selected
measured time series data, a number of residuals as show in
Table 15‑2 are estimated. The calculated residuals
are used for plotting against time or observed data. A histogram of
the standardized residuals is also made available for evaluating
the model fit.
Table 15‑2 –
Types of Residuals
No.
|
Criteria
|
Equation
|
1
|
Residuals
|
|
2
|
Absolute Residual
|
|
3
|
Squared Residual
|
|
4
|
Relative Residual
|
|
5
|
Absolute Relative Residual
|
|
6
|
Squared Relative Residual
|
|
7
|
Standardized Residuals
|
|
Histogram of Standardized Residuals
A histogram of the
standardized residuals are plotted to check the assumption that the
errors are normally distributed. The histogram of standardized
residual may also help to inspect if the residuals are randomly
scattered about zero and to approximately 95 percent of the
residuals lie between +2 and -2. A histogram produced for the BOD
dataset is as shown in Figure 15‑5.
Residual Plots against
Measured Data
The estimated
residuals can be plotted against the measured data point to see
trends and biases in model outputs. A typical output of this type
of plot is as shown in Figure 15‑6.
Residuals Plots against
Time
For identifying
trends in estimated residuals, the time plots of the residuals may
be useful. The model predictability may be affected some specific
time based events and these plots may be of help to identify time
periods of special events (rain fall, plant maintenance etc.) where
the plant operation conditions are not captured adequately in the
model. A typical output for this type of plot is as shown in
Figure 15‑7.
Figure 15‑5
- Histogram of Standardized Residuals
Figure 15‑6
- Residuals Plotted against the Observed Data
Figure 15‑7
- Residuals Plotted against Simulation Time
This feature allows the user to set the speed of the simulation to
real time or to a multiple of real time. This is useful in on-line
SCADA applications or for operator training in which GPS‑X acts as
a virtual plant.
The parameters for this feature
can be accessed at the bottom of the
General Data > System > Parameters > Simulation
Tools Settings form (Figure 15‑8), and select
the More… button in the On-line Operation
sub-section. The Real Time Synchronized Mode sub-section
contains the parameter that will switch the real time clock
ON or OFF. When the real time clock is ON, the
simulation speed is equal to a multiple of real time. The
multiplication factor used is specified by the Real Time
Accelerator Factor. (i.e. setting the accelerator factor to 24
causes one simulation day to take one real hour). Note, the real
time clock feature forces the Communication interval to be
equal to one second.
Figure 15‑8
- Real Time Clock Parameters
The steady-state solver used in GPS-X is a robust routine based on
a direct search algorithm (that is, no gradients are used).
Sometimes, the steady-state convergence appears slow or diverges
due to a problem in the way that the model is set up. For example,
if the underflow rate from a settler is too low, that settler will
begin to fill up with solids. Since the concentration ranges of
solids in the ten layers become large, the changes in these layers
between loops of the steady-state solver also become large which
may cause slow convergence or divergence.
In all cases of poor convergence, the user should closely examine
all the unit processes for proper specification. The easiest way to
do this is to examine the dsum# variable associated with
each unit process where # represents the overflow stream
label. You can do this by issuing the display command in the
Command line of the Simulation Control window. The
sum of the individual dsum values equals the
dsum value displayed
in the Log window. Having done that, there are a number of
parameters associated with the steady-state solver that can be
fine-tuned.
The forms shown in Figure 15‑9 and Figure
15‑10 are accessed by selecting
General Data > System > Parameters > Steady-State
Solver Settings. The parameters in these forms are
described below.
The first menu item, number
of retries on iteration indicates the number of times the
steady-state solver will try to reach convergence if it fails.
The error limit on individual variables is a tolerance below
which the steady-state ignores a variable. If the steady-state
solver is making changes to a state variable that become smaller
than this tolerance, the solver will ignore the variable, assuming
it is at steady-state.
When the solver achieves a sum of state variable derivatives below
the iteration termination criteria, a steady-state
convergence is triggered. The default value is 10.0, but in some
systems it may need to be increased when in single precision.
The contract constant and expand constant refer to
the step size made by the steady-state solver between iterations. A
larger step is taken if an improvement is found, while a smaller
step is taken if no improvement is found. The performance of the
steady-state solver can be greatly affected by these parameters.
The expansion factor is limited to the maximum step size in one
iteration. This maximum step size is dampened as the
steady-state solution is reached. The damping factor on final
approach will control the damping effect. A zero value means no
damping; while a value of 1.0 means that the maximum step size is
limited to one tenth of the maximum step size.
The initial step
size that the steady-state solver takes is governed by the
initial perturbation parameter. This number is normally set
smaller than the maximum step size to prevent a poor initial guess
by the solver.
Figure 15‑9
- Steady State Parameters
Figure 15‑10
- More Steady-State Parameters
The next four items
shown in these forms deal with other termination criteria and
output control. The status of the solver is printed at a frequency
controlled by the convergence output interval, while the
steady-state loop counter initial value specifies which
iteration loop will be first printed to the simulation Log
window. The maximum number of iterations limits the
steady-state solver to attempting only this many iterations. The
loop counter is reset after each retry. Another steady-state
termination criterion is the maximum number of unsuccessful
iterations which will cause the steady-state solver to
terminate if an improvement in the value of the sum of the
derivatives is not made within this number of loops.
The trim parameters shown here refer to the
output of the ACSL trim function. This is an alternative
steady-state solver, which uses the Jacobian matrix for accelerated
searches. However, this gradient type routine was not found to be
very robust with the large models encountered within GPS-X.
The Advanced
Steady-State Parameters sub-section contains parameters that are
used to enhance and refine the contraction process for each state
variable. The maximum concentration bound for each state variable
are found in the More… button menu.
The numerical issues discussed in this chapter are important
because the success of a simulation depends on the input data and
the models and on the numerical solver used to do the
calculations.
Any modeller using dynamic simulation should have an understanding
of the underlying numerical methods used to solve the equations to
properly interpret the results. Understanding the numerical methods
used will help to identify problems that are numerical in
nature.
The choice of which numerical solver to use is important. There are
several numerical integration methods available in GPS-X. The
numerical solver can be set in the Simulation Control window
or can be saved into a scenario or the layout by setting it
in
General Data > System > Input Parameters > Dynamic
Solver Settings and scrolling down to the Integration
Settings sub-section (Figure 15‑11).
Generally, the default
integration method, Runge-Kutta-Fehlberg (2), works very well for
most models; however, there are situations where the solver may
need to be changed. In general the trade-off is between numerical
accuracy and simulation speed. The fixed step algorithms
(i.e. Euler, Runge-Kutta (1), and Runge-Kutta (2)) are faster
than the variable step algorithms (i.e. Adams-Moulton,
Runge-Kutta-Fehlberg (1), Runge-Kutta-Fehlberg (2), Gear's Stiff,
and Differential Algebraic Solver) but are not as accurate.
If the system is stiff
then you may need to use the Gear's Stiff algorithm. A stiff system
is one in which there are processes occurring at very different
time scales (e.g., a very fast process such as the transfer of
oxygen into a tank and a very slow reaction occurring at the same
time).
In addition to the
numerical solvers, the user can change some of the numerical
parameters that control the simulations. Changing the default
parameters however requires extreme caution because in some cases
it may introduce errors in the results. The numerical parameters
are accessed by making the following selections:
General Data > System > Parameters >
Numerical. The form is shown in Figure
15‑12.
These parameters
include bounds on flow, state variables, state derivatives,
parameters, exponentials, and volumes. The parameters for the
implicit solver (IMPL - used for solving implicit functions
evaluated as residuals, where the residuals are reduced to zero)
used in the exact code option can be set here. Some of the
parameters are discussed in more detail in the sub-sections
below.
Figure 15‑11
– Integration Methods
Figure 15‑12
- Numerical Parameters Form
Ignore Dilution
Parameters
If the liquid
volume in a tank falls below the ignore dilution rate below this
volume parameter value, then the dilution rate (defined as the
inverse of the hydraulic residence time) will not be used to
calculate the concentrations in the tank as would normally be done.
In doing so, the speed of simulation is greatly increased. The same
principle is applied to the settlers and SBRs where the dilution
rate in a particular layer is ignored if the height of the layer is
smaller than the ignore dilution rate below this layer
thickness parameter value.
Protection against
Division by Zero
The parameter protect against division by zero is fixed and cannot
be changed by the user. This parameter is used in all calculations
involving division. It is used to ensure that the denominator in an
expression is never equal to zero. For example, the DO switching
function, used in many rate equations (shown below), includes the
protection against division by zero parameter.
Equation 15.1
Speed Sub-Section
The parameters in the Speed sub-section of the
Numerical form concern the pump flow rates used in all the
objects. Since the switching of pumps on and off introduces sharp
changes in the flows, the numerical integration routine may require
a very small step size to reduce the numerical error. In some
cases, these small step sizes will result in a very slow
simulation. If the smooth pump discharge at
discontinuitieslogical is turned ON (default is
OFF), then the pumped flow rate is ramped up to the
specified amount as follows:
Equation 15.2
where:
Q = pumped
flow rate (m 3/d),
par = smoothing
factor,
t
= smoothing period (d).
The smoothing function will only be applied to a user-specified
percentage change (smooth at flow changes larger than). This
allows the user the flexibility of smoothing only larger changes in
a pump flow rate. The smoothing function will be applied to
all pump flows in a layout including underflows, control splitters,
etc. See Figure 15‑13 for an illustration of how the
smoothing function affects the pumped flow.
Figure 15‑13
- Smoothing Function
Although GPS-X
allows the user to build very large and complex layouts, there are
some restrictions. First, the simulation language used, ACSL,
currently has a limit on the number of discrete blocks of code that
can be specified. What this means to the user is that the number of
automatic controllers for a given layout cannot exceed 25. Since a
number of objects have built-in controllers (e.g., settler has a
controller for both the pumped flow and underflow), the user can
quickly use all the allowable discrete code blocks. To circumvent
this problem, the user can select either the Big option,
which combines all the automatic controller code into one discrete
block, or the Big+ option, which disables all controllers.
In selecting the Big option, the user can no longer specify
the sampling time for each individual controller loop. One sampling
interval will override the rest. This sampling interval
(controller sampling time) is found under General Data
> System > Parameters > Miscellaneous.
Robust vs. Fast
Models
A trade-off exists between the speed of the simulation and the
robustness and accuracy of the model under different conditions
(e.g. sudden flow discontinuities or tanks emptying completely).
GPS-X is configured such that the average user with a typical
continuous flow activated sludge plant need not worry about the
accuracy and the speed of the simulation. However, when dealing
with more complex processes it may be necessary for the user to
switch to a different integration algorithm or change the values of
some numerical parameters to improve the accuracy or speed of the
simulation. In general, robustness and accuracy are more important
than simulation speed.
If you suspect that the simulation results are not accurate, you
should first check that the physical dimensions of the process
objects and the values of the process flows are reasonable. A
numerical value may have been entered incorrectly.
If you are confident that the model parameters are reasonable then
the next step is to determine whether a modelling simplification or
numerical problem is leading to inaccurate results. Some helpful
information is listed below:
·
If you suspect that the simulation results are not accurate, you
may wish to try the Exact code option. (Build >
Code menu item). The Quick option (default) results
in a faster simulation but is not as accurate when recycle flows
are present. When the Quick option is used the recycle flows
at each time step are simply the values taken from the previous
time step. When the Exact option is used the recycle flows
are calculated exactly using an implicit nonlinear equation solver.
In most situations using the Quick option is acceptable,
because the integration time steps are small and recycle flows
usually don't change dramatically. In an application with a number
of recycle flows and sudden extreme dynamic changes this
simplification may cause some problems.
·
The choice of integration algorithm (IALG) can have a significant effect on the output of
the model. For a detailed description of the available integration
algorithms consult the ACSL Reference Manual (contact
Hydromantis for details). Generally, only variable step algorithms
(i.e. Adams-Moulton, Runge-Kutta-Fehlberg(1),
Runge-Kutta-Fehlberg(2), Gear's Stiff, and Differential Algebraic
Solver) are acceptable for the simulation of highly dynamic
systems. Try the Gear's Stiff algorithm if you suspect that your
system is stiff (unless you are simulating an SBR or trickling
filter, in which case only the Runge-Kutta-Fehlberg algorithms are
allowed).
The Gear's Stiff routine is a self-tuning algorithm that adapts
itself to rapidly changing derivatives - in some cases though it
might require excessive simulation time to cross a discontinuity in
the model. The Runge-Kutta-Fehlberg(1) algorithm is exactly the
opposite - it is very forgiving for discontinuities, but
experiences difficulties with steeper problems. A good compromise
between these two solvers is the Adams-Moulton algorithm. It is
very robust and does not give up easily but it is slower. The
Runge-Kutta-Fehlberg(2) algorithm is almost identical to
Runge-Kutta-Fehlberg(1), but is more robust.
·
It is possible to monitor the step sizes used by the integration
algorithm during a run. You can either print the values to the
Log window by entering the following command at the command
line in the Simulation Control window: output
truecssitg or place the variable truecssitgon an output
graph. You can select truecssitgfor display in the
General Data > System > Output Variables > General
program variables form (it is labeled as Average integration
step size). The average integration step size
(truecssitg) is calculated over each communication
interval. Start your simulation using a small
communication interval (~0.01 or less). Sharp drops in
truecssitg usually signal a discontinuity. Using the above
method, the Sum of absolute values of the
derivatives (dsum) can also be checked for
sharp changes.
·
If your simulation is slow you can find which variable makes the
simulation slow by selecting the Fullinfo option in
the Options>Set up>Information menu of the
Simulation Controlwindow, and running the
simulation using either the dams-Moulton or Gear’s Stiff
integration algorithms. The error summary displayed at the
end of the run in the Log window will point to the variable
that controls the step size.
·
The integration of systems with low DO (dissolved oxygen)
concentrations (in the range of a few dozen micrograms) is time
consuming because the integration algorithm has to cut back on the
step size or risk running into negative DO values.
·
Sometimes the integration algorithms have difficulty at the start
of a simulation. In such cases the initial concentrations of the
state variables may be inappropriate, resulting in large initial
derivative values. In this case the variable step integration
algorithms keep cutting back on the step size to go over this
initial bump. It may be helpful to use the steady-state solver to
determine the initial conditions.
·
Simulation speed depends strongly on the values of the derivatives
of the state variables. You can reduce the absolute value of the
lower and upper bounds on the derivatives to speed up the
simulation of discontinuities by eliminating sharp peaks. These
bounds can be found in the
General Data > System >
Parameters > Dynamic Solver Settings form. The
best values to use depend on the system in question. Try plotting
the DO derivative (found in the Process Data menu for each object),
because this is the most sensitive state variable in most cases.
Get the maximum value during the run and then select the bounds to
be 50 percent of the highest value. Changing the bounds from the
default values (-1.e33 and 1.e33) can restrict the accuracy of the
integration.
·
Sudden start-up or switching off of pumps (influent or other) can
cause discontinuities in the derivatives, and as a consequence, a
delay in the processing of the simulation. To prevent this, a
smooth pump option is provided in the General Data > System > Parameters > Dynamic
Solver Settings form, which starts and stops pumps
smoothly. This is strictly intended as a numerical feature and does
not try to simulate real pump discharges during transient periods.
By default, the smoothing is not used. If you experience problems
when pumps are turned off or on, try using smoothing.
·
Check the Log window for
any messages printed during the run.